Color changing effects with cross-halftone prints

Abstract

We propose new methods for creating color varying prints with classical cyan, magenta, yellow inks on a metallic specularly reflecting or on a white diffusely reflecting substrate. We use a special cross-line halftone with optimized surface coverages of the inks to create cross-halftone prints whose colors change when rotating the print in-plane under specular reflection. These prints enable viewing at the same location a first image with a first set of colors and upon in-plane rotation of the print or displacement of the observer, a similar or a different image whose parts are colored with a different set of colors. Applications comprise counterfeit prevention, art, advertisement, decoration, exhibitions and surprising displays in amusement parks.

Claims

1. A method for embedding two input color images on the same area of a cross-halftone print whose cross-halftones comprise two intersecting sets of line halftones, the first input color image being viewable in non-rotated viewing mode and the second input color image being viewable in rotated viewing mode, the method comprising a preparation and a printing phase, where the preparation phase comprises the steps of: (i) establishing a color prediction model predicting the colors of said cross-halftones in non-rotated and rotated viewing modes as a function of ink surface coverages of said intersecting line halftones; (ii) creating a multi-dimensional desired color table mapping a first desired color in non-rotated viewing mode and a second desired color in rotated viewing mode to the ink surface coverages of the cross-halftone by minimizing with said color prediction model a difference metric between desired and achieved colors in non-rotated viewing mode and desired and achieved color in rotated viewing mode; (iii) computing at least one sub-gamut containing colors that are printable and that provide freedom for color changes from sub-gamut mapped colors of said first input image to sub-gamut mapped colors of said second input image; where the printing phase comprises the steps of (a) performing the gamut mapping of colors from said first input color image into said sub-gamut and of colors from said second input color image into said sub-gamut; (b) deducing, for each pair of colors from the first and second gamut mapped images, by accessing the multi-dimensional desired color table from step (ii) the surface coverages of the inks for creating the cross-halftoned ink separation layers; (c) with said surface coverages of the inks, creating by halftoning the cross-halftoned ink separation layers; (d) printing said cross-halftoned ink separation layers on a transparent layer that is superposed with a substrate layer selected from the set of metallic specularly reflecting layer and diffusely reflecting layer, thereby obtaining said cross-halftone print; and where directional optical dot gain causes colors produced by one of said sets of line halftones oriented perpendicularly in respect to the incident light to be substantially different from colors produced by the same set of line halftones oriented parallel to the incident light.

2. The method of claim 1, where in both non-rotated and rotated viewing modes, light illuminates the print in front of an observer, where the non-rotated viewing mode is achieved without rotation of the print and the rotated viewing mode is achieved by rotating the print in-plane by a quarter turn, where said directional optical dot gain occurs when light rays hit one of said sets of line halftones at a substantially perpendicular angle.

3. The method of claim 1, where for the substrate layer made of the diffusely reflecting layer the first color image is viewed by illuminating the print from the front of the observer and the second color image is viewed by illuminating the print from the side of the observer.

4. The method of claim 1, where after step (i) of establishing the color prediction model, a multi-dimensional cross-halftone surface coverage to color table is created by establishing with said color prediction model the correspondence between the ink surface coverages of said cross-halftones and the resulting colors viewable in non-rotated and rotated viewing modes, where after step (iii) of computing said sub-gamut, colors sampling the space of input colors are mapped to said sub-gamut and form a sub-gamut mapping table, where step (a) of the printing phase is performed with said sub-gamut mapping table, where the inks are the cyan (c), magenta (m), yellow (y) inks, where the intersecting sets of line halftones forming the cross-halftones are line halftones substantially parallel to the incident light and line halftones substantially transverse to the incident light, where the color prediction model predicting the colors in non-rotated and rotated viewing modes is a 6D ink cross-halftone color prediction model, and where said multi-dimensional cross-halftone surface coverage to color table is a 6D cross-halftone surface coverage to color table.

5. The method of claim 1, where in case that the second input color image differs mainly from the first input color image by hue variations, said sub-gamut is a hue variation sub-gamut, where in case that the second input color image is a grayscale image, said sub-gamut is a decolorization sub-gamut and where in case that the second input color image is independent of the first color image, said sub-gamut is an independent color variation sub-gamut.

6. The method of claim 4, where said 6D ink cross-halftone color prediction model accounts for directional dot gain of the transverse line halftones by relying on an ink-spreading enhanced cellular Yule-Nielsen modified spectral Neugebauer (IS-CYNSN) prediction model specially adapted to predict the reflectance of line-on-line ink halftones, said IS-CYNSN prediction model being calibrated by spectral reflectance measurements of selected line halftones.

7. The method of claim 4, where minimizing with said 6D ink cross-halftone color prediction model a difference metric between the desired and the achieved colors in non-rotated viewing mode and the desired and the achieved colors in rotated viewing modes is performed with the formula:
{cmy.sub.P,cmy.sub.T}=arg min(ΔE.sub.94(Lab(cmy.sub.P,cmy.sub.T),Lab.sub.InNR)+ΔE.sub.94(Lab(cmy.sub.T,cmy.sub.P),Lab.sub.InR))
subject to
0≦{c.sub.P,m.sub.P,y.sub.P,c.sub.T,m.sub.T,y.sub.T}≦1 where cmy.sub.P stands for the nominal surface coverages (c.sub.P, m.sub.P, y.sub.P) of the parallel line halftone, cmy.sub.T stands for the nominal surface coverages (c.sub.T, m.sub.T, y.sub.T) of the transverse line halftone, where Lab(cmy.sub.P, cmy.sub.T) is the color predicted in non-rotated mode by the 6D ink halftone color prediction model, where Lab(cmy.sub.T, cmy.sub.P) is the color predicted in rotated mode by the 6D ink halftone color prediction model, where Lab.sub.inNR and Lab.sub.inR are the desired colors in non-rotated and rotated modes, respectively, and where ΔE.sub.94 is a standard color difference formula for calculating color differences in CIELAB space.

8. The method of claim 1, where the first and second images are images selected from the set of color picture, grayscale picture, image showing a text message, image showing an alphanumeric code, image showing a barcode, image showing a data matrix code and image showing a QR code.

9. The method of claim 1, where the cross-halftone print embedding the first and second images is part of a valuable item such as a passport, an identity card, a credit card, a debit card, a personal entry card, a medical card, a diploma, a driving license, a hunting license, a fishing license, an original business document, a classified document, a shipping document, an event ticket, a transportation ticket, a product label, a tax label, a home care product, a beauty product, a health care product, a food product, a beverage product, a pharmaceutical, a drug, an electronic equipment, a valuable metallic component, an aerospace component, an automotive component, a sporting product, a fashion product, a jewellery item, a watch, a compact disk, a DVD, and where the presence of said first and second image on said same area protects said valuable item against counterfeits.

10. The method of claim 1, where the cross-halftone printed transparent layer is superposed with said substrate layer by a process selected from pasting, lamination, polymerization, shrink sleeve.

11. The method of claim 1, where said substrate layer is formed beneath said printed transparent layer by a particle deposition process selected from vaporization, sputtering, spin-coating, spraying, plating, coating and roll-to-roll coating.

12. The method of claim 1, where the cross-halftone print embedding the first and second images is part of one of the following elements: an advertisement, a piece of art, a floor, a wall, a decoration, a surprising display.

13. A computing system for embedding by cross-halftones two input color images on the same area of a print made of a plastic layer superposed with a specularly or diffusely reflecting substrate, the first input color image being viewable under directed light in non-rotated viewing mode and the second input color image being viewable under directed light in rotated viewing mode, said computing system comprising a CPU, memory, disks, a network interface and callable software functions, which in a preparation phase are operable: (i) to calibrate a prediction model predicting under directed light colors in non-rotated and rotated viewing modes as a function of ink surface coverages of a cross-halftone formed by intersecting line halftones; (ii) to create with said prediction model a multi-dimensional cross-halftone surface coverage to color table by establishing the correspondence between the ink surface coverages and the resulting colors viewable in non-rotated and rotated viewing modes; (iii) to create a multi-dimensional desired color table mapping a desired color in non-rotated viewing mode and a desired color in rotated viewing mode to ink surface coverages of the cross-halftone by minimizing with the prediction model a difference metric between the desired and achieved colors in non-rotated viewing mode and the desired and achieved color in rotated viewing modes; (iv) to compute at least one sub-gamut under specular viewing conditions containing colors that are printable and that provides freedom for color changes from gamut mapped colors of the first input image to gamut mapped colors of the second input image; (v) to gamut map input colors sampling the space of colors to said sub-gamut and to form with the resulting gamut mapped sampled input colors a corresponding sub-gamut mapping table; and which in a printing phase are operable (a) to perform with said sub-gamut mapping table the gamut mapping of colors from said first input color image into said sub-gamut and of colors from said second input color image into said sub-gamut, thereby obtaining a first gamut mapped image viewable in non-rotated viewing mode and a second gamut mapped image viewable in rotated viewing mode; (b) to deduce, for each pair of colors from the first and second gamut mapped images, by accessing the multi-dimensional desired color table, the surface coverages of the inks for creating cross-halftoned ink separation layers; (c) with said surface coverages of the inks, to create by halftoning the cross-halftoned ink separation layers; and (d) to send said cross-halftoned ink separation layers to the printer.

14. The computing system of claim 13, where the inks are the cyan (c), magenta (m), yellow (y) inks, where the intersecting line halftones forming the cross-halftone are line halftones substantially parallel to the incident light and line halftones substantially transverse to the incident light, where the model predicting the colors in non-rotated and rotated viewing modes is a 6D ink cross-halftone color prediction model, and where said multi-dimensional cross-halftone surface coverage to color table is a 6D cross-halftone surface coverage to color table.

15. The computing system of claim 13, where the first and second images are independent images selected from the set of color picture, grayscale picture, image showing a text message, image showing an alphanumeric code, image showing a barcode, image showing a data matrix code and image showing a QR code.

16. A print with colored cross-halftones superposed with a substrate selected from the set of specularly reflecting metallic substrate and diffusely reflecting substrate, said print embedding two different images on the same area and showing to an observer the first of said two images in non-rotated viewing mode and the second of said two images in rotated viewing mode, where the non-rotated viewing mode is achieved by illuminating the print from light located in front of the observer and the rotated viewing mode is achieved by one of the following actions: rotating the print in plane by a quarter turn, displacement of the observer around the print, in case where said substrate is the diffusely reflecting substrate, illuminating the print from the side of the observer, and where the main differences between the first image and the second image are from the group of hue variations, decolorizations, recolorizations, and independent color variations.

17. The print of claim 16, whose cross-halftones are formed by two line halftones, one in transverse orientation and one in parallel orientation in respect to the plane of incidence determined by the normal of the print and the illumination direction.

18. The cross-halftone print of claim 16 which is part of one of the following elements: an advertisement, a piece of art, a floor, a wall, a decoration, a surprising display.

19. The cross-halftone print of claim 16, where the first image shows an element selected from the set of picture, text message, image showing a barcode, image showing a data matrix code and image showing a QR code, and where the second image shows a different element, selected from the same set.

20. The cross-halftone print of claim 16, where one of the images contains an information selected from the set of text message, barcode, data matrix code and QR-code, where a computing device acquires a photograph of said information, where said computing device in case of a text message applies optical character recognition techniques and in case of a barcode, data matrix code and QR-code applies code decoding techniques to obtain the information, and where said computing device authenticates the information, possibly by interacting with a remotely located computing server.

21. The print of claim 16, which is part of a valuable item such as a passport, an identity card, a credit card, a debit card, a personal entry card, a medical card, a diploma, a driving license, a hunting license, a fishing license, an original business document, a classified document, a shipping document, and event ticket, a transportation ticket, a product label, a tax label, a home care product, a beauty product, a health care product, a food product, a beverage product, a pharmaceutical, a drug, an electronic equipment, a valuable metallic component, an aerospace component, an automotive component, a sporting product, a fashion product, a jewellery item, a watch, a compact disk, a DVD, where the presence of said first and second image at the same location protects said valuable items against counterfeits.

22. The computing system of claim 13, where in the case of the print made of a superposition of the plastic layer with the diffusely reflecting substrate, the non rotated viewing mode is achieved by illuminating the print in front of an observer and the rotated viewing mode is achieved by illuminating the print from the side of the observer.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0019] FIG. 1 shows an example of a setup to observe a print under directed light, with a light table 101, the print 100, the observer 102 and the in-plane rotation of the print 103;

[0020] FIG. 2 shows an example of a setup to observe under directed light a cross-halftone print 210 when walking from a first position 212 to a second position 213, seeing at the first position mainly light reflected by a first array of lines 215 in transverse mode and at the second position mainly light reflected by the second array of lines 214 in transverse mode;

[0021] FIG. 3 shows an example of a cross-halftone 301, enlarged 302 and 303, made of intersecting horizontal (305, transverse) and vertical (306, parallel) line halftone structures with the ink layers printed “line-on-line”, resulting in colorants red (R) 310, cyan (C) 311, blue (B) 312, and black (K) 313;

[0022] FIG. 4A shows a segment of a line halftone located perpendicularly to the incident light I.sub.w (400), i.e. in transverse mode;

[0023] FIG. 4B shows a segment of a line halftone located parallel to the incident light I.sub.w (400), i.e. in parallel mode;

[0024] FIG. 4C represents a cross-section of the print which shows that in transverse mode, light attenuated by the transparent ink line halftone exits also from non-inked regions (430);

[0025] FIG. 5 shows for line halftones superposed with a metallic substrate the lightness L* and chroma C* of the cyan line halftones with nominal surface coverages of 0%, 25%, 50%, 75% and 100% measured in parallel and transverse modes (i.e. parallel to the incident light and perpendicular to the incident light) where the differences in lightness and chroma between the parallel and transverse modes demonstrate the impact of the large dot gain in transverse mode;

[0026] FIG. 6 shows for cross-halftones superposed with a metallic substrate the color behavior upon rotation of a cross-halftone formed by two inks, one with lines laid out in transverse mode and the second with lines in parallel mode, where for example, “y.sub.t50c.sub.p50” means a halftone with 50% surface coverage of yellow in transverse mode and 50% surface coverage of cyan in parallel mode and where the other end of the dashed line is the color after 90° turn in-plane rotation;

[0027] FIG. 7 shows for cross-halftones superposed with a metallic substrate the color behavior of 50% surface coverage halftones formed by 3 inks yielding a very large hue shift, where the marker close to the label shows the color of the print before rotation and the other end of the dotted line shows the color after rotation;

[0028] FIG. 8 shows one of the 8 sub-cubes obtained by the combinations of 0%, 50% and 100% surface coverages of the three cyan (C), magenta (M) and yellow (Y) inks;

[0029] FIG. 9A represents a cross-section and FIG. 9B a top view of a line-on-line printed cyan, magenta, yellow line halftone, for surface coverages c≦m≦y, where the & sign indicates the superposition of inks, C, M, Y represent ink line surfaces of cyan, magenta and yellow, respectively, where W represents the unprinted substrate and where R and K represent the red and black colorants, respectively, formed by superpositions of inks Y & M and Y & M & C, respectively;

[0030] FIG. 10 shows a table with the line-on-line equations describing colorant area coverages a.sub.w, a.sub.c, a.sub.m, a.sub.y, a.sub.r, a.sub.g, a.sub.b, a.sub.k as a function of ink surface coverages ĉ, {circumflex over (m)}, ŷ, where indices w, c, m, y, r, g, b, k stand for white, cyan, magenta, yellow, red, green, blue, and black, respectively;

[0031] FIG. 11 shows an ink spreading curve mapping normalized nominal ink surface coverages to normalized effective ink surface coverages;

[0032] FIG. 12 shows the ink spreading enhanced cellular Yule-Nielsen modified spectral Neugebauer model (IS-CYNSN), for line-on-line ink halftones;

[0033] FIGS. 13A and 13B show a parallel and a transverse line halftone, respectively, with a first ink 131 on which a second ink 132 and a third ink 133 are superposed to form a line-on-line halftone;

[0034] FIGS. 14A and 14B show a cross-halftone screen element with intersecting parallel and transverse line halftones comprising the substrate (white, area Ni)), a parallel line halftone (area iP) formed by ink surface coverages (c.sub.P, m.sub.P, y.sub.P) a transverse line halftone (area iT) formed by ink surface coverages (c.sub.T, m.sub.T, y.sub.T), and an inked halftone center (area iPT) located at the intersection of the transverse and parallel line halftones;

[0035] FIG. 15 shows a table giving the prediction accuracy of the line-on-line cyan, magenta and yellow IS-CYNSN model for reflectances measured in specular mode (25°:25°, expressed as average, 95% quantile and maximal ΔE.sub.94 color differences between halftone patch predictions and halftone patch measurements, for test sets of 125 halftones representing all combinations of 0, 25%, 50%, 75% and 100% c, m, y ink surface coverages;

[0036] FIG. 16 shows the prediction accuracy for the 6 ink spectral model predicting the reflectance of cross-halftones, expressed as average, 95% quantile and maximal ΔE.sub.94 color difference, with test samples forming all c, m, y, ink combinations with values of 0%, 40% and 80% for the parallel and transverse components of the cross-halftone;

[0037] FIG. 17 shows desired pairs (“circles”) and achievable pairs (“triangles”) of colors incorporating lightness shifts upon in-plane rotation of the print by 90°;

[0038] FIGS. 18A and 18B show desired 45° and 90° hue shifts, respectively, with “circles” (a.sub.1-b.sub.1, a.sub.2-b.sub.2, a.sub.3-b.sub.3, a.sub.2-c.sub.2, a.sub.3-c.sub.3) giving pairs of desired initial and hue-shifted colors and “triangles” or “squares” giving the pairs of corresponding achieved initial and hue shifted colors, the shifted colors appearing when rotating the print by 90°;

[0039] FIG. 19 shows initial chroma values (“circles”) to be reduced to a very low chroma (C*≈0) upon in-plane rotation, i.e. reduced to achromatic colors, as well as the achievable printable colors on the non-rotated and rotated print that are shown as triangles;

[0040] FIG. 20A shows the print gamut 2001 and the sub-gamuts allowing for specific hue shifts Δh.sub.ab, where any color within each sub-gamut can achieve the corresponding hue shift while keeping the achieved non-rotated color equal to the desired non-rotated color;

[0041] FIG. 20B shows the print gamut 2001 and the corresponding decolorization sub-gamut 2010, allowing chroma shifts from chromatic colors inside the gamut to achromatic colors and vice-versa;

[0042] FIG. 21 shows projections on the L*a*, L*b* and a*b* planes of the print gamut 2101 and of the independent color changing sub-gamut 2102 allowing to produce a rotated color image that is independent of the non-rotated image;

[0043] FIG. 22 shows the color reproduction workflow specific for creating images incorporating the color changing effects that are viewable under specular reflection;

[0044] FIG. 23A shows schematically the non-rotated butterfly with red 2305 foreground colors and green 2301 background colors and FIG. 23B the rotated hue modified butterfly image with cyan 2306 foreground colors and magenta 2302 background colors, viewed under directed light;

[0045] FIG. 24A shows schematically the non-rotated initial “gray” 2401 and FIG. 24B the rotated “chromatic” umbrella image, with magenta 2402, cyan 2403, yellow 2404, orange 2405 and red 2406 sections viewed under directed light;

[0046] FIG. 25A shows schematically under directed light a non-rotated initial identity card image, where the front view of a colored face 2501 appears on the left and the side view of a face, in gray 2502, appears on the right and FIG. 25B shows the same identity card image, but 90° in-plane rotated, where the front view of the face appears now in gray 2511 and where gray text related to the ID card holder appears on the right side 2512;

[0047] FIG. 26A shows the images of two messages covering the same location of a cross-halftone print, one appearing in non-rotated mode (2601, schematically represented by gray letters) and the other that appears in rotated mode (2602, schematically represented by black letters) and FIG. 26B shows how the first message appear, with a first color or gray value 2610 for the background elements and a second color or gray value 2611 for the foreground elements (letters), and FIG. 26C shows how the second image appears upon 90° rotation of the print, with the background having a third color or gray value 2620 and the foreground having a fourth color or gray value 2621, where all 4 color or gray values may be different;

[0048] FIG. 27A show the overlay of two QR-code messages within one color changing image, where one message (shown in gray) appears in non-rotated mode as shown in FIG. 27B and the other message (shown in black) appears in rotated mode as shown in FIG. 27C;

[0049] FIGS. 28A and 28B show two independent color pictures covering the same area, where under directed light one picture with low chroma brown 2801 and green 2802 colors appears in non-rotated mode, as shown in FIG. 28A, and the other picture with low chroma blue 2811 and red 2812 colors appears in rotated mode as shown in FIG. 28B;

[0050] FIG. 29 shows a computing system for embedding color images on cross-halftone prints that change colors upon in-plane rotation;

[0051] FIGS. 30A and 30B show a cross-halftone 3004 superposed with a white diffusing substrate 3005 with horizontal 3001 and vertical 3002 halftone lines, superposed with a white diffusing layer, where the observer 3003 sees a first image under front illumination 3000 and the second different image under side illumination 3010;

[0052] FIGS. 31A and 31B show a more detailed view of an instance of a cross-halftone formed by parallel ink lines 3101, superposed with a diffusing substrate 3100, where in FIG. 31A, light comes from the front 3112, traverses inked 3116 or non-inked 3118 regions, is diffusely reflected by the diffuse substrate (3114, 3015) and exits (3117, 3119) without additional colorization due to dot gain and where in FIG. 31B, light comes from the side 3110, traverses in the present example an inked 3102 region, is diffusely reflected 3103 towards non-inked and inked regions by the diffuse substrate 3100 and partly exits through a non-inked region 3104 in the direction of the observer 3111, being colored due to the dot gain;

[0053] FIG. 32 shows lightness L* and chroma C* of the cyan line halftones printed on plastic superposed with a white diffusing substrate, with nominal surface coverages of 0%, 25%, 50%, 75% and 100% measured in parallel and transverse modes (i.e. parallel to the incident light and perpendicular to the incident light) where the differences in lightness and chroma between the parallel and transverse modes demonstrate the presence of an important dot gain in transverse mode, both for the (45°:45°) and for the (65°:0°) measuring geometries;

[0054] FIG. 33 shows the color changes obtainable by a cross-halftone superposed with a diffusing substrate formed by two inks, one with lines laid out in transverse mode and the second with lines in parallel mode, where for example, “y.sub.t50c.sub.p50” means a halftone with 50% surface coverage of yellow in transverse mode and 50% surface coverage of cyan in parallel mode and where the other end of the dashed line is the color after a 90° in-plane rotation of the print, or in case of the (65°:0°) measuring geometry, possibly after 90° turn in-plane rotation of the light source;

[0055] FIG. 34A shows for the (45°:45°) geometry the color behavior of 50% surface coverage halftones formed by 3 ink cross-halftones superposed on a diffusing substrate yielding large hue shifts, where the marker close to the label shows the color of the print before rotation and the other end of the dotted line shows the color after rotation; and

[0056] FIG. 34B shows for the same cross-halftone print as in FIG. 34A, but for the (65°:0°) geometry, very large hue shifts.

DETAILED DESCRIPTION OF THE INVENTION

[0057] We observed that non-isotropic halftones such as line halftones printed on a flat transparent medium of a certain thickness located on top of a specularly reflecting metallic or of a diffusely reflecting white substrate changes its color upon in-plane azimuthal rotation of the print. Line halftones superposed at a certain elevation (e.g. 30 μm to 150 μm) on top of metal or on top of a white diffusing layer induce an anisotropic dot gain effect. There is a strong dot gain when the halftone line screen is perpendicular to the incident light and a low dot gain when the halftone line screen has the same orientation as the incident light, see FIGS. 4A and 4B. These observations led us to the idea of creating color halftones made of two superposed structures, one structure being formed by lines having a transverse orientation in respect to the incident light and a second structure formed by lines having a parallel orientation in respect to the incident light. These two structures form a cross-halftone (FIG. 3). Each line structure comprises freely chosen surface coverages of the superposed cyan, magenta and yellow inks. Due to the large dot gain of colored lines in transverse orientation, rotating the cross-halftone print by approximately 90° in its own plane (azimuthal rotation) modifies the hues and chroma of the cross-halftones and to some extent also their lightnesses.

[0058] We build upon the hue and chroma changing capabilities of cross-halftone prints viewed under directed light to create color images that keep their overall shape, but change their colors when being rotated azimuthally by approximately 90°. The idea is to let a designer take one color image and to allow him to recolor this color image with new hues or with modified chroma values, using a tool such as Adobe Photoshop. This modified image represents then the second image. The user may also specify two independent images. For each pair of initial and modified colors, the proposed method finds the solution that is the closest to the desire of the user, i.e. the surface coverages of the cross-halftone inks that yield a color as close as possible to the initial color for the non-rotated print and a color as close as possible to the modified color for the rotated print.

[0059] The computation framework for obtaining the surface coverages of the inks yielding the desired pair of initial and modified colors relies on a spectral prediction model. This spectral prediction model has 6 input parameters, namely the cyan (c), magenta (m) and yellow (y) surface coverages of inks for the horizontal and for the vertical line halftone structures. This spectral prediction model predicts the color of the cross-halftone print at its initial orientation as well as the color of the print after 90° in-plane rotation. By minimizing a cost function relying on the prediction model for computing the difference between a desired pair of colors and the corresponding achievable pair of colors, we obtain the surface coverages of the three inks (c.sub.T, m.sub.T, and y.sub.T) for a line halftone in transverse orientation and the surface coverages of the three inks (c.sub.P, m.sub.P, and y.sub.P) for a line halftone in parallel orientation.

[0060] We consider in the present work two viewing conditions, a viewing condition under directed light, where the print is at its initial position and a second viewing condition under directed light where the print is azimuthally rotated by approximately 90°. For these two specific viewing conditions, we propose a new model making accurate color predictions as a function of the 6 surface coverages of the inks.

Vocabulary

[0061] Classical inks are substantially transparent chromatic or achromatic inks that are in general use for printing on paper, i.e. primarily the cyan, magenta and yellow inks as well as further colored inks, such as light cyan, light magenta, red, green, blue, orange, pink, violet, or achromatic inks such as gray and black inks. Transparent inks are inks that absorb part of the light in some wavelength range and transmit the remaining light to the next layer. The substrate layer can be either a flat metallic layer that reflects the incident light primarily in the specular direction or a diffusely reflecting layer that reflects from the cross-halftone print a large part of the incident light towards the observer. In a preferred embodiment, the diffusely reflecting layer is a white layer. However, a colored diffusely reflecting layer could also be used to create cross-halftone prints.

[0062] The terms “cross-line halftone” and “cross-halftone” are used interchangeably.

[0063] The terms “specular viewing conditions”, “specular observation conditions”, “specular viewing mode”, “cross-halftone print viewed from the front of an observer under directed light” indicate that the print is viewed by having the incident light and the light reflected by the cross-halftone print towards the observer in the incident plane.

[0064] The “incident plane”, “incident light plane” or “plane of incidence” is the plane formed by the an incident light ray (FIG. 1, 105) originating from a light source 101 and hitting the print 100 and the normal of the print 106. In case of a metallic specularly reflecting substrate, the incident plane comprises also the light reflected in the direction of the observer's eyes 102.

[0065] The term directed light indicates that light illuminates the print mainly from a certain direction, i.e. from a solid angle that is only a fraction of the hemisphere. It also means that environmental light illuminating the cross-halftone print is considerably less intense compared with the directed light.

[0066] In non-rotated viewing mode, the print appears as originally printed, with the horizontal halftone lines, called “transverse lines”, perpendicular to the incident plane illuminating the print in front of the observer and the original vertical halftone lines, called “parallel lines”, lying parallel to the incident plane. In a preferred cross-halftone screen, the set of parallel and the set of transverse halftone lines intersect each other and are orthogonal one to another. They form a cross-line halftone.

[0067] In rotated viewing mode, the following variants are possible:

[0068] the print is rotated in-plane by a quarter turn (i.e. approximately 90° see FIG. 1, 103),

[0069] the observer changes its position in respect to the print (e.g. in FIG. 2, from 212 to 213),

[0070] the light source and the observer are on the same platform and rotate around of the print,

[0071] in case of a white diffusing substrate, the print and the observer remain in their original positions and the light source is rotated azimuthally by 90° or another light source is activated that illuminates the print from the side of the observer.

[0072] In non-rotated viewing mode, a first image appears to the observer. In rotated viewing mode, the second image appears at the same position as the first image and the first image is not visible any more.

[0073] A “non-rotated” color is a color that appears in non-rotated viewing mode. A “rotated” color is a color that appears in rotated viewing mode, e.g. after quarter turn in-plane rotation of the cross-halftone print.

[0074] The term “an image that changes its color” is equivalent to “two images that are embedded within the same area” or “two images that cover the same area”. The first image is the non-rotated image that appears before the change of colors and the second image is the rotated image after the change of colors. The term “color changing effect” has the same meaning as “color varying effect”.

[0075] The term “90° in plane rotation”, “quarter turn” or “azimuthal 90° rotation” are used interchangeably. The terms “non-rotated mode” and rotated mode” are identical to “non-rotated viewing mode” and “rotated viewing mode”.

[0076] “Color change” is used as general term that can mean a hue shift, a chroma change, e.g. from a chromatic color to an achromatic (gray) color, or a general change in hue, chroma and lightness.

[0077] The expression “inks printed in superposition with a substrate” or “cross-halftones printed in superposition with a substrate” means that there is a gap between the ink halftone layer (FIG. 4C, 431) and the substrate 433. This gap is generally formed by a transparent plastic layer of a certain thickness 432. But it may be formed by other transparent means such as glass or air.

[0078] A metallic substrate may be embodied by a metalized film. Such a metalized film is a plastic layer that is metalized by a deposition process (e.g. vapor deposition) that brings metallic particles onto the surface of the plastic layer.

[0079] A diffusely reflecting substrate may be embodied by paper, by plastic incorporating diffusely reflecting particles or by a thin layer of deposited white diffusing particles.

[0080] Cross-halftones printed on a plastic layer superposed with a diffusely reflecting substrate may be called “diffusely reflecting cross-halftone print” and cross-halftones printed on a plastic layer superposed with a specularly reflecting metallic substrate may be called “specularly reflecting metallic halftone print”.

[0081] In the present invention, the term “color” comprises also the achromatic colors (gray values or gray levels), i.e. a color image may also be a grayscale image.

[0082] Regarding the measurement geometry, for the angle of the incident light, “incident angle” is used interchangeably with “zenithal angle” or simply “zenith angle”. The measurement geometry described as (45°:45°) means: the incident light illuminates the sample at an angle of 45 degrees in respect to the normal and the light reflected from the sample is captured at an angle of 45 degrees in respect to the normal.

Viewing and Illumination Conditions

[0083] One may view the print incorporating the color changing effects preferably under directed light in front of a window. By remaining under directed light and rotating the print in-plane by 90 degrees (azimuthal rotation), one observes the color changes. Instead of looking at the print in front of a window, one may create a setup where a flat uniformly illuminating light table (FIG. 1, 101) is placed in a dark room. The light table may be placed at an oblique orientation and the print 100 placed so as to view it under specular viewing conditions (FIG. 1A). By rotating 103 the print 100 in plane and keeping it in specular viewing condition, one may observe the color changing effect.

[0084] One may also create a large setup (FIG. 2), where the print 210 is illuminated 211 from one side, for example by a large window. Then, when walking along that print at a certain distance from it, the observer may see from position 212 the print first in transverse mode in respect to line halftone 215 and in parallel mode in respect to line halftone 214 and then by walking to the next position 213 in parallel mode in respect to line halftone 215 and in transverse mode in respect to line halftone 214. Such a setup may be used for showing posters, for publicity or for entertainment.

[0085] A further way of observing cross-halftone prints having a diffusely reflecting substrate consists in observing the cross-halftone print by illuminating it from the front of the observer (FIG. 30A) and then by illuminating it from the side of the observer (FIG. 30B).

Experimental Setup

[0086] Let us give an example explaining how color shifting cross-halftones are printed and measured. We use the Epson StylusPro WT7900 commercially available printer for printing cyan, magenta and yellow ink halftones on a ink-absorbing plastic layer stacked with a metalized film at a resolution of 1440 dpi. The ink halftone screen that we generate consists of horizontal and vertical lines. Horizontal and vertical lines are formed of variable surface coverages of the cyan, magenta and yellow ink halftones superposed line-on-line (FIG. 3).

[0087] Both horizontal and vertical line halftones are printed at e.g. at 100 lpi or 150 lpi. To create the cross-halftone ink layer, we create in the computing system's memory the horizontal and vertical line halftone layers for the cyan, magenta and yellow inks, according to their surface coverages. The cross-halftoned ink layer is obtained by printing a pixel of a given ink if it belongs to either the corresponding horizontal or vertical line halftone ink layer. We create the cross-halftone ink layers for the cyan, magenta and yellow inks and send them to the printer.

[0088] The reflectances of the printed cross-halftones are measured with a custom-built spectrophotometer. This spectrophotometer is designed to mimic the viewing conditions when the sample is viewed under specular reflection from a window or from a small area light source. In the case of cross-halftone prints superposed with a metallic substrate, the samples were measured at specular illumination and reflection angles centered around the angle 25° in respect to the normal, i.e. according to the geometry (25°:25°). In the case of prints superposed with a white diffusing substrate, the reflectances were measured either at the (45°:45°) geometry or at the (60°:0°) geometry. For cross-halftones superposed with a metallic substrate, the light source is a 150 W Halogen Dolan-Jenner DC-950H DC-Regulated Fiber Optic Illuminator connected through a fiber optic light guide to a 4.25″×3.37″ illuminator from Edmund Optics providing a highly diffuse and even light distribution. For cross-halftones on a white diffusing substrate, the incident light exits from an optical fiber having a 600 μm diameter and a 12.5° half-angle cone aperture. In both cases, the reflected light is collected by an optical fiber having a 600 μm diameter and a 12.5° half-angle cone aperture mounted on a graded rotational stage. The optical fiber is connected to a Maya 2000 Pro back-thinned CCD spectrophotometer from Ocean Optics. Reflectances are calculated by dividing the captured irradiance of the considered print by the captured irradiance of the unprinted surface.

[0089] CIELAB colors are deduced from the obtained reflectances by first converting the reflectances to the CIE-XYZ space under the D65 illuminant and for the CIE 1931 Standard Observer, with the unprinted silver substrate set as Y=100. Then the CIE-XYZ coordinates are converted to the CIELAB space by selecting the brightest patch under specular viewing conditions, i.e. the unprinted surface, to be the white reference point. Color differences are expressed with the ΔE.sub.94 metric. Color differences below 1 are not perceptible and between 1 and 2 are barely perceptible by the human eye, see Sharma, G. Color fundamentals for digital imaging, in Digital Color Imaging Handbook (G. Sharma Ed.), Chapter 1, CRC Press, 1-43 (2003), incorporated by reference and hereinafter referenced as [Sharma 2003].

Directional Optical Dot Gain

[0090] In halftone prints, the dot gain expresses the difference between the effective surface coverage of an ink and the nominal surface coverage that the printer is asked to print. In halftone prints on paper, the effective surface coverage is determined by the actual area of the printed dot and also by the lateral propagation of light within the substrate due to subsurface scattering, see the article by R. Balasubramanian, Optimization of the spectral Neugebauer model for printer characterization, Journal of Electronic Imaging 8, 2, 156-166 (1999), incorporated by reference and hereinafter referenced as [Balasubramanian 1999].

[0091] Halftones printed on a transparent layer located at a certain height over a specularly or a diffusely reflecting substrate show under specular reflection directional optical dot gain. The dot gain effect depends on the anisotropy of the halftone shape in respect to the plane formed by the illumination orientations and the normal of the print. The impact of the azimuthal orientation of the print on the directional dot gain is maximized by superposing parallel and transverse halftone lines. Transverse halftone lines are perpendicular to the incoming light rays. They contribute more to the overall color, since they have a large directional dot gain (FIG. 4A). Halftone lines parallel to the incoming rays have a negligible directional dot gain (FIG. 4B). Example FIG. 4A shows schematically that the reflected light in transverse mode is of color magenta (m) 402, magenta (m) 401 and magenta (m) 403, i.e. the light traversing areas 405 and 407 close to the inked area as well as the inked area 406 becomes colored according to the color of the inked area. In contrast, FIG. 4B shows schematically that the light reflected in parallel mode is composed of white (w) 410, magenta (m) 411 and white (w) 412, i.e. the reflected light traversing areas close to the inked area remains uncolored.

[0092] The print is composed of the ink halftone layer (FIG. 4C, 431), a transparent ink attracting polymer 432 of a certain thickness (e.g. 30 μm, 50 μm, 100 μm, or 150 μm), a substrate made of a reflecting specular metallic or a white diffusing layer 433 and the background support 434. The ink halftone slightly diffuses the incoming light. Due to the reflection of light on the substrate and to Fresnel reflections at the polymer-air interface, multiple reflections of light occur. At each of these reflections, part of the light exits the print and part of it is internally reflected. In transverse mode, due to the height of the ink attracting polymer, light attenuated by the ink halftone may exit the print at non-inked locations. This results in the important directional optical dot gain that appears when the halftone lines are laid out in transverse mode, i.e. perpendicularly to the incident light.

[0093] FIG. 5 shows in the case of a metallic substrate the effect of the directional optical dot gain on the color measured with our custom spectrophotometer. We measured under specular reflection (geometry: 25°:25°) cyan halftones lines printed in superposition with a metalized film laid out both in parallel mode and in transverse mode with nominal surface coverages of 0%, 25%, 50%, 75% and 100%. The directional optical dot gain has a strong impact on the lightness and chroma of the samples (FIG. 5). In transverse mode, the directional dot gain is maximized and yields darker and higher chroma samples compared to the same samples measured in parallel mode, where the directional dot gain is minimal. At 0% and 100% surface coverage, the orientation is irrelevant.

[0094] In order to have a better view of the color shifts that can be achieved with two or three inks, let us examine several special cases of interest. FIGS. 6 and 7 show in the case of a cross-halftone superposed with a metallic substrate the colors of cross-halftones for different inks that have each 50% nominal surface coverages. In FIG. 6, a first ink line halftone is laid out in transverse mode (index t) and the second ink line halftone is laid out in parallel mode (index p). In FIG. 7, the second and third ink are superposed on the line halftone in parallel mode. After 90° in-plane rotation of the print, the ink line halftone laid out in transverse mode is in parallel mode and the ink(s) line halftone laid out in parallel mode are in transverse mode. The corresponding color shifts are shown by the dotted lines. The color of the ink lines in transverse mode dominates.

The Ink Spreading Enhanced Cellular Yule-Nielsen Modified Spectral Neugebauer (IS-CYNSN) Model for Line-On-Line Halftones

[0095] The 6D color prediction model established for predicting the colors of cross-halftones relies partly on the cellular Yule-Nielsen modified spectral Neugebauer (IS-CYNSN) model for line-on-line halftones.

[0096] The IS-CYNSN reflectance prediction model relies on the same principles as the Yule-Nielsen modified Spectral Neugebauer model (YNSN), see [Balasubramanian 1999]. In the YNSN model, the predicted spectral reflectance of a halftone is

[00001]$\begin{array}{cc}R\left(\lambda \right)={\left(\underset{i}{\overset{}{.Math.}}.Math.a_i.Math.{R_i\left(\lambda \right)}^{1/n}\right)}^n& \left(1\right)\end{array}$

where R.sub.i(λ) is the measured spectral reflectance of the i.sup.th fulltone colorant (also called Neugebauer primary), a.sub.i is its effective area coverage, and n is an empirical scalar coefficient related to the halftone screen frequency and halftone dot profile. In the YNSN model, colorants are formed by all possible superpositions of fulltone inks and paper white. In case of the three inks cyan (c), magenta (m) and yellow (y), there are 2.sup.3=8 fulltone colorants: white (no inks), cyan, magenta, yellow, blue (cyan & magenta), red (yellow & magenta), green (cyan & yellow) and black (cyan & magenta & yellow). The “&” sign indicates a superposition of the fulltone inks.

[0097] In the IS-CYNSN model, the three-dimensional ink space (c, m, y) is subdivided into 8 sub-cubes whose vertices are given by colorants formed by combinations of c, m, y nominal surface coverages of 0, 0.5 and 1 (fulltone). For three inks, there are 3.sup.3=27 colorants (primaries). One of the sub-cubes is shown in FIG. 8.

[0098] In line-on-line halftone screens, each ink line is printed in exact superposition with the other ink lines (FIGS. 9A and 9B). The colorants that are present within a given superposition of cyan, magenta and yellow lines depend on their respective surface coverages. For example, in FIG. 9, with an ink dot surface relationship c≦m≦y, the area coverages a.sub.k, a.sub.r, a.sub.y, a.sub.w of the colorants black (superposition of cyan, magenta and yellow), red (superposition of yellow and magenta), yellow (yellow printed alone) and white, respectively, are the following: [0099] a.sub.k=c; surface coverage of the colorant black [0100] a.sub.r=m−c; surface coverage of the colorant red [0101] a.sub.y=y−m; surface coverage of the colorant yellow [0102] a.sub.w=l−y; surface coverage of the colorant white.
Within each sub-cube, the input ink surface coverages (c, m, y) are normalized (ĉ, {circumflex over (m)}, ŷ):

[00002]$\begin{array}{cc}\hat{c}=\frac{c-c_l}{c_h-c_l}.Math..Math.\hat{m}=\frac{m-m_l}{m_h-m_l}.Math..Math.\hat{y}=\frac{y-y_l}{y_h-y_l}& \left(2\right)\end{array}$

where c.sub.l, m.sub.l, and y.sub.l indicate the lowest surface coverages (0 or 0.5) within the considered sub-cube, and c.sub.h, m.sub.h, and y.sub.h indicate the highest surface coverages (0.5 or 1) within the considered sub-cube.

[0103] In order to predict the spectral reflectance for the line-on-line halftones, the normalized effective area coverages a.sub.i of the colorants in function of the normalized ink surface coverages are calculated according to the entries of the table shown in FIG. 10.

[0104] Due to the spreading of ink dots on the underlying substrate and inks, the normalized effective surface coverages c′, m′, y′ differ from the normalized nominal surface coverages ĉ, {circumflex over (m)}, ŷ, see publication by R. Rossier and R. D. Hersch, 2010. “Introducing ink spreading within the cellular Yule-Nielsen modified Neugebauer model”, in Proc. IS&T 18th Color Imaging Conference, 295-300, herein incorporated by reference and referenced as [Rossier and Hersch 2010].

[0105] FIG. 11 shows an ink spreading curve mapping normalized nominal ink surface coverages to normalized effective surface coverages û.fwdarw.f(û). We recover ink spreading curves by printing the color sample located at the center of each sub-cube. Normalized effective dot sizes C.sub.0.5′, m.sub.0.5′, y.sub.0.5′ at the center of each sub-cube are fitted by minimizing the square difference between the measured reflectance and the reflectance predicted according to Equation (1), by performing the conversion between normalized ink surface coverages and colorant area coverages according to the table of FIG. 10. Within each sub-cube, three ink spreading curves mapping nominal to effective surface coverages are formed by linearly interpolating between points (0,0) (0.5, u.sub.0.5′) and (1,1), where the u.sub.0.5′ is a place holder for c.sub.0.5′, m.sub.0.5′, y.sub.0.5′.

[0106] FIG. 12 shows the ink-spreading enhanced cellular Yule Nielsen prediction model, relying on line-on-line ink halftone superpositions. For given nominal ink surface coverages c, m and y, the spectral reflectance R(c, m, y) is predicted by locating the corresponding sub-cube j, calculating the normalized ink surface coverages ĉ, {circumflex over (m)}, ŷ, obtaining effective ink surface coverages c.sub.j′, m.sub.j′, y.sub.j′ from the ink spreading curves, calculating the effective colorant area coverages a.sub.ji′ from the line-on-line equations (table entries in FIG. 10) and inserting them into the YNSN equation (1). For model calibration, in the case of three inks, 27 patches are required for measuring the colorant reflectances and 8 additional patches are required for establishing the ink spreading curves. In total, 35 printed patches are needed to calibrate the 3 ink line-on-line halftone IS-CYNSN model. The IS-CYNSN model for line-on-line halftones is similar to the dot-on-dot halftone model described by [Balasubramanian 1999], with ink-spreading carried out as described by [Rossier and Hersch 2010].

6D Spectral Prediction Model for Cross-Halftones

[0107] A spectral prediction model creates the relation between the amounts of inks, i.e. the surface coverages of the ink halftones and the observed color by relying on predicted color halftone reflectances. In order to create interesting color changing effects with cross-halftones, we developed a special spectral model predicting the reflectance and therefore the color obtained when printing with the six ink surface coverages (c.sub.T, m.sub.T, y.sub.T, c.sub.P, m.sub.P, y.sub.P) of the superposed transverse and parallel line halftones. This model takes into account the directional optical dot gain. It enables predicting the reflectance of a halftone for the two considered azimuthal angles (0° and 90°) of the print.

[0108] As explain in FIGS. 14A and 14B, the parallel and the transverse halftone lines divide each cross-halftone screen element into four areas: non-inked part (Ni), inked part with lines parallel to the orientation of the incident light (iP), inked part with transverse lines that are perpendicular to the orientation of the incident light (iT) and inked part formed by the intersection of the parallel and transverse lines (iPT).

[0109] Let us show how to create the model predicting the colors of the 6 ink cross-halftone before and after azimuthal rotation of the print. In a first step, we establish prediction models for predicting under directed light separately the reflectances of the parallel and transverse line halftones. In a second step, we deduce the reflectances of each of the parts (iP, iT, iTP) shown in FIG. 14B. In a third step, we assemble these parts by weighting their reflectances according to their surface coverages, applying the Yule-Nielsen modified spectral Neugebauer model described by formula (1).

[0110] Let us first predict separately the reflectance R.sub.HalftoneP(λ) of the parallel line halftone (FIG. 13A) as a function of the parallel line ink surface coverages (c.sub.P, m.sub.P, y.sub.P) as well as the reflectance R.sub.HalftoneT(λ) of the transverse line halftone (FIG. 13B) as a function of the transverse line ink surface coverages (c.sub.T, m.sub.T, y.sub.T). We rely on two instances of the “Ink-spreading enhanced cellular Yule-Nielsen modified spectral Neugebauer model (IS-CYNSN)” for cyan, magenta and yellow line-on-line halftones, with one level of subdivision (see above, Section “The ink spreading enhanced cellular Yule-Nielsen modified spectral Neugebauer (IS-CYNSN) model for line-on-line halftones”). The first IS-CYNSN model predicts the reflectances of parallel line halftones and the second IS-CYNSN model predicts the reflectances of transverse line halftones.

[0111] In each IS-CYNSN model, the 3D c, m, y ink surface coverage space is divided into 2.sup.3=8 subcubes with vertices at 0, 50% and 100% nominal surface coverages of each of the inks. There are therefore 3.sup.3=27 primary reflectances. The ink spreading enhanced Yule-Nielsen modified spectral Neugebauer model (IS-YNSN) is separately applied on each subcube. To account for ink spreading, the central point within each sub-cube is printed and measured [Rossier and Hersch 2010]. The three effective cyan, magenta and yellow ink surface coverages are fitted to minimize the square difference between the measured spectral reflectance and the reflectance predicted by the YNSN model. With the fitted effective surface coverages of the 3 inks at the 8 combinations of 25% and 75% nominal surface coverages, the model constructs by linear interpolation in each sub-cube ink spreading curves mapping nominal to effective ink surface coverages (FIG. 11). In the case of parallel line halftones, the mapping is from (ĉ.sub.P, {circumflex over (m)}.sub.P, ŷ.sub.P) to (c.sub.P′, m.sub.P′, y.sub.P′) and in the case of transverse line halftones, the mapping is from (ĉ.sub.T, {circumflex over (m)}.sub.T, ŷ.sub.T) to (c.sub.T′, m.sub.T′, y.sub.T′).

[0112] The table of FIG. 15 shows that the two resulting line-on-line IS-CYNSN models are capable of accurately predicting reflectances under directed light both for the parallel and for the transverse line halftones superposed on a metallic substrate. The predictions of the transverse line halftones are slightly less accurate, presumably due to the important dot gain.

[0113] In order to predict the reflectances of the different parts of a cross-halftone, we first derive the relative nominal and effective widths of the inked areas separately for the parallel (FIG. 13A) and the transverse (FIG. 13B) halftones. Since the inks are printed as “line-on-line”, the width of the inked area corresponds to the largest ink surface coverage, i.e. for the nominal widths

a.sub.iP=max(c.sub.P, m.sub.P, y.sub.P)

a.sub.iT=max(c.sub.T, m.sub.T, y.sub.T)   (3a)

and for the effective widths

a.sub.iP′=max(c.sub.P′, m.sub.P′, y.sub.P′)

a.sub.iT′=max(c.sub.T′, m.sub.T′, y.sub.T′)   (3b)

Let us derive the reflectances of the inked areas present within the parallel (FIG. 13A) and within the transverse (FIG. 13B) line halftones.

[0114] Each line halftone of reflectance R.sub.halftone(P|T)(λ) is formed by a non-inked part of effective area (1−a.sub.i(P|T)′) having the measured reflectance of the unprinted surface R.sub.M(λ) and of an inked part of effective area a.sub.i(P|T)′ having reflectance R.sub.i(P|T)(λ), where index (P|T) stands either for parallel (P) or transverse (T) line halftones. According to the YNSN model (formula (1)), the overall reflectance of a line halftone is

R.sub.halftone(P|T).sup.1/n(c.sub.(P|T), m.sub.(P|T), y.sub.(P|T))=a.sub.i(P|T)′R.sub.i(P|T).sup.1/n+(1−a.sub.i(P|T)′)R.sub.M.sup.1/n   (4)

[0115] where n is the empirical scalar accounting for part of the optical dot gain.

[0116] By predicting with the IS-CYNSN model the reflectance R.sub.halftone(P|T )(λ) of the line halftone in function of the nominal ink surface coverages (c.sub.(P|T), m.sub.(P|T), y.sub.(P|T)), we can deduce the reflectance R.sub.i(P|T)(λ) of the inked area of either a parallel or a transverse halftone as follows:

[00003]$\begin{array}{cc}{R}_{i\left(P.Math.\backslash .Math.T\right)}^{1/n}\left(c_{\left(P.Math.\backslash .Math.T\right)},m_{\left(P.Math.\backslash .Math.T\right)},y_{\left(P.Math.\backslash .Math.T\right)}\right)=\frac{R_{\mathrm{halftone}\left(P.Math.\backslash .Math.T\right)}^{1/n}\left(c_{\left(P.Math.\backslash .Math.T\right)},m_{\left(P.Math.\backslash .Math.T\right)},y_{\left(P.Math.\backslash .Math.T\right)}\right)-\left(1-a_{i\left(P.Math.\backslash .Math.T\right)}^{\prime }\right).Math.R_M^{1/n}}{a_{i\left(P.Math.\backslash .Math.T\right)}^{\prime }}& \left(5\right)\end{array}$

In order to estimate the reflectance of the intersection area (iPT in FIG. 14B), we first calculate the relative surface coverages of the 3 inks (c.sub.iPT, m.sub.iPT, y.sub.iPT) within the inked intersection. We assume that in the intersection area iPT, the reflectance is approximately the average of the parallel and transverse area reflectances. We therefore calculate according to the IS-CYNSN model separately the reflectances of the inked parallel and transverse line halftones with these relative surface coverages and finally take the average of the two as the reflectance R.sub.iPT(λ) of the inked intersection.

[0117] The relative nominal surface coverages (c.sub.iPT, m.sub.iPT, y.sub.iPT) of each ink at the inked intersection of the lines are obtained by adding the corresponding nominal ink surface coverages u.sub.P and u.sub.T of the parallel and transverse ink lines, each one normalized in respect to the nominal inked areas a.sub.iP and a.sub.iT (FIGS. 13A and 13B) respectively, and subtracting once the normalized surface coverage of their intersection:

[00004]$\begin{array}{cc}{u}_{\mathrm{iPT}}=\frac{u_P}{a_{\mathrm{iP}}}+\frac{u_T}{a_{\mathrm{iT}}}-\frac{u_P.Math.u_T}{a_{\mathrm{iP}}.Math.a_{\mathrm{iT}}}& \left(6\right)\end{array}$

where u stands for the nominal surface coverage of inks c, m or y.

[0118] The reflectance R.sub.iPT(λ) of the inked halftone part at the intersection of the two sets of inked lines is assumed to be the average of the parallel and transversal line reflectances, R.sub.halftoneP(λ) and R.sub.halftoneT(λ) respectively, predicted with the relative nominal surface coverages c.sub.iPT, m.sub.iPT, y.sub.iPT calculated according to Eq. (6):

[00005]$\begin{array}{cc}{R}_{\mathrm{iPT}}=\frac{R_{\mathrm{halftoneP}}\left(c_{\mathrm{iPT}},m_{\mathrm{iPT}},y_{\mathrm{iPT}}\right)+R_{\mathrm{halftoneT}}\left(c_{\mathrm{iPT}},m_{\mathrm{iPT}},y_{\mathrm{iPT}}\right)}{2}& \left(7\right)\end{array}$

According to the YNSN model, we obtain the overall reflectance of the screen element (FIG. 5c), by taking the reflectances of the individual parts and weighting them according to their effective surface coverages. The areas iP, iT, iPT, Ni within the halftone (FIG. 14B) have the effective surface coverages a.sub.IP′.Math.(1−a.sub.IT′), a.sub.IT′.Math.(1−a.sub.IP′), a.sub.IP′.Math.a.sub.IT′, (1−a.sub.IP′).Math.(1−a.sub.IT′), respectively. We obtain:

R.sub.total(λ).sup.1/n=a.sub.iP′.Math.(1−a.sub.iT′)R.sub.iP(λ).sup.1/n+a.sub.iT′.Math.(1−a.sub.iP′)R.sub.iT(λ).sup.1/n+a.sub.iP′.Math.a.sub.iT′.Math.R.sub.iPT(λ).sup.1/n+(1−a.sub.iP′)(1−a.sub.iT′)R.sub.M(λ).sup.1/n   (8)

where R.sub.iP(λ) is the reflectance of the inked halftone lines parallel to the incident light with low directional dot gain, R.sub.iT(λ) is the reflectance of the inked transverse halftone lines with high directional dot gain, R.sub.iPT(λ) is the reflectance of the inked intersection area, R.sub.M(λ) is the measured reflectance of the unprinted surface.

[0119] The predicted spectral reflectance (and color) of a cross-halftone R.sub.total(c.sub.P, m.sub.P, y.sub.P, c.sub.T, m.sub.T, y.sub.T) is calculated by first predicting the reflectances of the parallel and of the transverse line halftones, i.e. R.sub.halftoneP(c.sub.P, m.sub.P, y.sub.P) and R.sub.halftoneT(c.sub.T, m.sub.T, y.sub.T) with the ink-spreading enhanced cellular Yule-Nielsen modified Neugebauer model (IS-CYNSN). Then, with Eqs. (3a) and (3b), one computes with the surface coverages of the inks both the nominal and effective sizes of the parallel and of the transverse line areas within a screen element, i.e. a.sub.iP, a.sub.iP′, a.sub.iT, a.sub.iT′. This enables computing according to Eq. (5) the reflectances of the inked areas R.sub.iP and R.sub.iT of the line halftones. For the intersection area a.sub.iP.Math.a.sub.iT, the relative surface coverages c.sub.iPT, m.sub.iPT, y.sub.iPT of the superposition of parallel and transverse lines are computed according to Eq. (6). The reflectance R.sub.iPT of the intersection area is calculated according to Eq. (7). Finally the reflectance of the whole halftone R.sub.total is assembled from the reflectances of the areas according to Eq. (8).

[0120] When rotating the print by a 90° azimuthal rotation, the parallel and transverse ink surface coverages get exchanged. The resulting halftone reflectance is calculated in the same way as above, but with R.sub.total(c.sub.T, m.sub.T, y.sub.T, c.sub.P, m.sub.P, y.sub.P).

[0121] In order to test the prediction accuracy of the overall model, we printed halftones at all nominal c, m, y ink combinations of 0%, 40% and 80% for both the parallel and the transverse ink surface coverages. This yields 3.sup.6=729 test halftone samples for the set of 6 inks (c.sub.P, m.sub.P, y.sub.P, c.sub.T, m.sub.T, y.sub.T). These test samples are completely distinct from the calibration samples and cover a large part of the reproducible color space. The two IS-CYNSN models for predicting the R.sub.halftoneP and R.sub.halftoneT reflectances require a total of 35*2−8=62 patches for their calibration (8 fulltone colorants are the same for parallel and transverse halftones and are measured only once). In all reflectance weighting equations, the n-value is set to 14. FIG. 16 shows the prediction accuracy for the proposed 6 ink cross-halftone spectral prediction model, for cross-halftones printed in superposition with a metallic substrate. The average prediction accuracy is approximately half the prediction accuracy of the single orientation parallel or transverse line halftones (FIG. 15). This prediction accuracy is good, since with a calibration set of only 62 patches, 729 colors are predicted.

Optimization Formula for a Desired Color Shift

[0122] Thanks to halftone screens incorporating parallel and transverse cyan, magenta and yellow lines of freely choosable width, one may create one color in non-rotated viewing mode and a second color in rotated viewing mode. The prediction model developed in the previous section enables, by varying the six ink surface coverages (c.sub.P, m.sub.P, y.sub.P, c.sub.T, m.sub.T, y.sub.T) to explore the space of possible pairs of colors that appear before and after 90° azimuthal rotation of the print.

[0123] Equation (9) yields the optimal surface coverages of the inks that minimize the sum of ΔE.sub.94 color distances between a pair of desired non-rotated and rotated colors and the pair of resulting printable non-rotated and rotated colors:

{cmy.sub.P,cmy.sub.T}=arg min(ΔE.sub.94(Lab(cmy.sub.P,cmy.sub.T),Lab.sub.InNR)+ΔE.sub.94(Lab(cmy.sub.T,cmy.sub.P),Lab.sub.InR))

Subject to

0≦{c.sub.P, m.sub.P, y.sub.P, c.sub.T, m.sub.T, y.sub.T}≦1   (9)

where cmy.sub.P stands for the nominal surface coverages (c.sub.P, m.sub.P, y.sub.P) of the parallel lines of the cross-halftone, cmy.sub.T stands for the nominal surface coverages (c.sub.T, m.sub.T, y.sub.T) of the transverse lines of the cross-halftone, where Lab(cmy.sub.P, cmy.sub.T) is the color predicted in non-rotated mode, where Lab(cmy.sub.T, cmy.sub.P) is the color predicted in rotated mode and where Lab.sub.inNR and Lab.sub.InR are the desired colors in non-rotated and rotated modes, respectively.

[0124] The color predictions Lab(cmy.sub.P, cmy.sub.T) and Lab(cmy.sub.T, cmy.sub.P) are obtained by calculating with Eq. (8) the reflectances R.sub.total(cmy.sub.P, cmy.sub.T) and R.sub.total(cmy.sub.T, cmy.sub.P). These reflectances are converted to CIE-XYZ tri-stimulus values and to CIELAB. The optimization of Eq. (9) is performed by applying the “Bound Optimization By Quadratic Approximation” method, see M. J. D. Powell 2009. The BOBYQA algorithm for bound constrained optimization without derivatives, Cambridge NA Report NA2009/06, University of Cambridge, UK.

[0125] In order to offer a high-speed access to the optimal surface coverages obtained by solving Eq. (9), we create a 6-dimensional desired color table comprising as entries the desired color coordinates Lab.sub.InNR of the non-rotated print and the desired color coordinates Lab.sub.InR of the rotated print. The L* coordinate is sampled from 20 to 100 in steps of 8, a* is sampled from −60 to 84 in steps of 6 and b* is sampled from −70 to 98 in steps of 7, providing a total of 11.Math.25.Math.25=6875 CIE L*a*b* coordinates. For all 6 coordinates, the table has 6875.sup.2=47′265′625 entries. At each entry, the optimal nominal surface coverages (c.sub.P, m.sub.P, y.sub.P, c.sub.T, m.sub.T, y.sub.T) of the cross-halftone are listed.

Exploring the Space of Hue, Chroma and Lightness Shifts

[0126] Let us explore how far it is possible to modify the hue, chroma and the lightness of colors when rotating the cross-halftone print superposed with a metallic substrate in-plane by 90° under directed light. Equation (9) enables specifying a desired color before rotation and a desired modified color after 90° in-plane rotation. The obtained surface coverages (c.sub.P, m.sub.P, y.sub.P, c.sub.T, m.sub.T, y.sub.T) yield the achievable color Lab(cmy.sub.P, cmy.sub.T) before rotation and the achievable color Lab(cmy.sub.T, cmy.sub.P) after rotation. FIG. 17 shows examples where the pairs of desired colors (“circles”) contain only a shift in lightness. One can observe that, if a large lightness difference is desired, the achievable lightness difference is moderate. In certain cases, the chroma also slightly changes (e.g. samples 1 and 3).

[0127] FIGS. 18A and 18B show examples of desired hue shifts (white “disks”). One can verify that desired and achievable hue shifts are close one to another when the chroma is not higher than C*=10 (samples a.sub.3-b.sub.3 and a.sub.3-c.sub.3). Desired hue shifts at chroma values 10<C*<20 are slightly attenuated (samples a.sub.2-b.sub.2 and a.sub.2-c.sub.2) and desired hue shifts at chroma values C*≧20 are considerably attenuated.

[0128] FIG. 19 shows examples of desired and achieved chroma shifts. Chroma shifts towards achromatic colors are always possible. Such chroma shifts are useful for transforming a color image into a grayscale image by rotating the print. Important chroma changes do not significantly affect the lightness.

[0129] Note that in all examples, the color changes can be inversed, e.g. from the shifted hue to the non-shifted hue and from achromatic colors to chromatic colors. To be more precise, exchanging Lab.sub.InNR and Lab.sub.InR in Eq. (9) yields the same set of surface coverages, but with the solutions cmy.sub.P and cmy.sub.T exchanged.

[0130] The set of colors achievable in rotated mode is obtained by first creating a cross-halftone surface coverage to color table mapping in steps of 5% the 6 surface coverages (c.sub.P, m.sub.P, y.sub.P, c.sub.T, m.sub.T, y.sub.T) to non-rotated and rotated colors. We associate to each of the 21.sup.6=8576612 entries of this cross-halftone color table both the corresponding non-rotated and rotated colors calculated according to the 6-ink spectral prediction model. We then search in this table all possible surface coverages in a small neighbourhood around the initial non-rotated color (0<{|ΔL*|, |Δa*|, |Δb*|}<1.5). With this set of surface coverages, we predict the corresponding colors in rotated mode.

[0131] We conclude that the potential for important lightness shifts is low, but that significant hue and chroma shifts are possible for colors that have medium to low chroma values. Pictures with medium and low chroma look generally less colorful than pictures with high chroma colors. Nevertheless, due to the Hunt effect which states that when the luminance of a given color increases, its perceived colorfulness also increases, samples of low chroma colors printed on metal and viewed under specular reflection appear to be colorful, see Hunt, R. W. G. Light and dark adaptation and the perception of color, J. Opt. Soc. Am. 42, 190-199 (1952).

[0132] Another important limitation of 90° cross-halftone rotation is that two colors having the same desired non-rotated color and different rotated colors may show a ghosting effect on the achieved non-rotated colors. This is due to the fact that the joint optimization of desired non-rotated and rotated colors according to formula (9) modifies the non-rotated colors by an amount that depends on the desired rotated colors. This can be deduced from FIG. 18A, where the achieved initial colors (e.g. the positions of “triangle” a.sub.1 and “square” a.sub.1) depend on the rotated colors (“circle” b.sub.1 and “circle” c.sub.1). However, for gamuts that are limited in their chroma and lightness values, the ghosting effects can be avoided (see Section “Print gamut and constrained color variation sub-gamuts”).

[0133] The presently analyzed color changing effects are, in addition to hue variations, primarily applied for hue changes, decolorization or independent color changes. Upon 90° in-plane rotation, image parts may take upon rotation hues that are different from the non-rotated hues, chromatic image parts may become achromatic, achromatic image parts may become chromatic, the lightness of images may undergo moderate changes or combined hue, chroma and lightness variations (called “color changes”) may be achieved.

Print Gamut and Constrained Color Variation Sub-Gamuts

[0134] Classic color reproduction workflows generally comprise a gamut-mapping step where input colors, e.g. sRGB display colors, are mapped into the gamut of the printable colors, e.g. the colors formed by the classic set of CMYK inks, see J. Morovic, J. Lammens, Color Management, in Colorimetry: Understanding the CIE system, (Ed. J. Schanda), Chapter 7, J. Wiley, 159-206 (2007).

[0135] We would like to create images that upon azimuthal rotation of the print by 90° provide color variations such as variations of hue, variations from chromatic to achromatic colors (decolorization) or combined variations of hue, chroma and to some extent lightness. For this purpose, we need to establish the subset of printable colors (sub-gamut) that comprise those colors, whose ink surface coverages provide the freedom to achieve the desired color variations, thereby eliminating the ghosting effects. The sub-gamut comprising colors providing the freedom to achieve any target hue from any starting hue is called the hue variation sub-gamut. The sub-gamut comprising all colors providing the freedom to achieve a full decolorization, i.e. a shift from chromatic colors to achromatic colors is called the decolorization sub-gamut. The sub-gamut comprising all colors providing the freedom to freely specify a color change, i.e. a color change incorporating a combination of hue, chroma and lightness variations is called either independent color variation sub-gamut or independent color change sub-gamut.

[0136] Color reproduction of pairs of input images incorporating hue variations, decolorization or independent color variations upon 90° rotation of the print requires projecting the colors present in the two input images, e.g. sRGB display colors, into the corresponding sub-gamut.

[0137] The experiments shown in the previous section teach that large hue changes can only be achieved with low chroma colors. Small hue changes however can be carried out with higher chroma colors. FIG. 20A shows sub-gamuts of colors supporting a given hue change, with the print gamut 2001, the Δh.sub.ab=20° hue variation sub-gamut 2002, the Δh.sub.ab=40° hue variation sub-gamut 2003, the Δh.sub.ab=90° hue variation sub-gamut 2004, and the Δh.sub.ab=170° hue variation sub-gamut 2005. The gamut and sub-gamut boundaries specify the achievable ranges of chroma and lightness. Note that Δh.sub.ab specifies the hue shift in terms of hue angle in the CIELAB color system.

[0138] Sub-gamuts of colors specifying those colors that after rotation show a hue shifted from the initial color by a given angle Δh.sub.ab are obtained by accessing the cross-halftone surface coverage to color table, see Section “Exploring the space of hue, chroma and lightness shifts” and for a given non-rotated color Lab.sub.NR, select those surface coverages that produce a non-rotated color within a window of e.g. (ΔL*, Δa*, Δb*)=(±1.5, ±1.5, ±1.5) of the given color. On these surface coverages, we verify if one of the surface coverages produces upon quarter turn rotation a “rotated” color that has a hue shift equal or larger than Δh.sub.ab. If this is the case, the corresponding “non-rotated” color Lab.sub.NR is said to belong to the sub-gamut that provides freedom for a hue shift of Δh.sub.ab. We repeat this procedure on all CIELAB colors that are inside the printable gamut, i.e. colors that are present in the cross-halftone color table. All colors that provide freedom for the given hue shift form the sub-gamut.

[0139] A similar procedure is applied to obtain the decolorization sub-gamut. On the surface coverages producing a non-rotated color within a window e.g. of (ΔL*, Δa*, Δb*)=(±1.5, ±1.5, ±1.5) of the given color, we verify if one of the surface coverages produces a decolorized color, i.e. a color having chroma lower than a given value, e.g. C*<3.

[0140] A similar procedure is also applied to obtain the independent color variation sub-gamut. Among the sets of surface coverages producing a non-rotated color within a window e.g. of (ΔL*, Δa*, Δb*)=(±1.5, ±1.5, ±1.5) of the given color, we verify if one set of surface coverages produces a hue change of at least 170°, if an other set of surface coverages provides the lightness of a pseudo-white, e.g. a lightness of L*≧70 and if a further set of surface coverages provides the lightness of a pseudo-black, e.g. a lightness of L*≦50. If these conditions are fulfilled, then the current color is part of the independent color variation sub-gamut. For such colors, cross-halftone surface coverages can be selected which may create upon quarter turn rotation a pseudo-white, a pseudo-black or any desired hue.

[0141] Let us hereinafter call the hue-shifted sub-gamut, the decolorization sub-gamut and the independent color variation sub-gamut by the global term “constrained color variation sub-gamuts”.

[0142] In a complex input image colorization task, upon rotation of the print, some parts can undergo an important hue shift while other parts remain with the same hue. Therefore, there is no necessity to reduce the chroma of all image parts by mapping their colors to the maximal Δh.sub.ab=180° hue variation sub-gamut. Instead, we propose to associate a value ρ to every image region that undergoes a hue shift, with 0≦ρ≦1. Value ρ indicates the relative proximity of the considered color between the maximal hue variation sub-gamut (ρ=1) and the full print gamut (ρ=0). Regions that remain with the same color after rotation of the print have a value of ρ=0.

[0143] In a real application, the designer may specify the image regions whose hues are to be changed. This may be carried out with a software package such as Photoshop. He shall specify on these regions the hue that should appear after rotation of the print. The software will then show a preview of the achieved initial and hue-shifted colors. The designer may then with a slider reduce the ρ value and therefore allow higher chroma values by tolerating, that the achieved non-rotated colors deviate towards the achieved modified hue.

[0144] Gamut-mapping to the target sub-gamut specified by the ρ value is performed by linearly interpolating between values obtained by gamut-mapping into the two sub-gamuts that have ρ values enclosing the desired p value. Pre-computed sub-gamuts are the sub-gamuts allowing a hue shift of Δh.sub.ab=0° (print gamut, ρ=0), Δh.sub.ab=45° (ρ=0.25), Δh.sub.ab=90° (ρ=0.5), Δh.sub.ab=135° (ρ=0.75), and Δh.sub.ab=180° (maximal hue-variation sub-gamut, ρ=1).

Synthesizing the Color Image with the Desired Hue, Chroma or Color Variations

[0145] The present color reproduction workflow consists of a preparation phase and of a printing phase. The preparation phase consists of characterizing the printer by calibrating the model predicting the reflection spectra of the cyan, magenta and yellow cross-line halftones, establishing the print gamut, the target hue-shifted sub-gamut, the maximal hue variation sub-gamut, the decolorization sub-gamut and the independent color changing sub-gamut, creating the tables mapping input sRGB values to these gamut and sub-gamuts, as well as creating the 6-dimensional desired color table describing the correspondence between the two input desired CIELAB values and the optimal surface coverages of the inks (FIG. 22, preparation phase, where LAB.sub.inNR stands for non-rotated desired color and LAB.sub.inR stands for rotated desired color).

[0146] The sub-gamuts associated to the hue shifts of 0°, 45°, 90°, 135° and to maximal hue variations of 180° are created by selecting the sets of achieved non-rotated colors allowing for these hue shifts and by computing the non-convex hull of these colors in the CIELAB space, see the article by Bernardini, F., Mittleman, J., Rushmeier H, Silva, C., and Taubin, G. The Ball-Pivoting Algorithm for Surface Re-construction, IEEE Trans. Vis. and Comp. Graph. 5, 4, 349-359 (1999), hereinafter referenced as [Bernardini et al. 1999]. FIG. 20A illustrates examples of sub-gamuts allowing hue shifts of Δh.sub.ab of 20° (2002), 40° (2003), 90° (2004) and 170° (2005).

[0147] A similar procedure is applied to obtain the decolorization sub-gamut (FIG. 20B, 2010 and the sub-gamut providing freely chosen color variations (FIG. 21, 2102). The print gamut (FIGS. 20A, 20B, 2001, FIG. 21, 2101) is obtained (i) by predicting with the 6D prediction model for cross-halftones the colors that are obtained in non-rotated mode when varying the surface coverages of the 6 inks and (ii) by computing the non-convex hull of these colors, in the same way as mentioned above.

[0148] The tables describing the gamut mappings are created by sampling the sRGB space by steps of 5%, converting sRGB values to CIELAB and mapping them separately onto the print gamut and onto the different sub-gamuts by applying a two-foci gamut-mapping algorithm, see J. Morovic, M. R. Luo, The fundamentals of gamut mapping: a survey, Journal of Imaging Science and Technology 45, 3, 283-290 (2001). The 6-dimensional desired color table describing the relationship between two CIELAB values and the 6 ink surface coverages is created as described at the end of Section “Optimization formula for a desired color shift”.

[0149] The printing phase consists in two input sRGB images specifying the colors before and after 90° in-plane rotation of the print as well as possibly a map with the p values. The second sRGB image differs from the first image by incorporating the desired hue shifts and/or the desired decolorization and/or the desired independent color changes. Using the previously established gamut-mapping tables, pairs of colors from the two images are mapped to the print gamut and to the sub-gamuts allowing for hue changes, decolorization or independent color changes. For hue shifts, parameter p defines by interpolation the colors of the gamut mapped RGB values, see Section “Print gamut and constrained color variation sub-gamuts”. The resulting pair of gamut mapped colors is then used to access the 6-dimensional desired color table yielding the 6 surface coverage values for creating the cross-halftone, 3 surface coverages for the parallel halftone lines and 3 surface coverages for the transverse halftone lines.

[0150] The cross-halftoned ink separation layers are conceived by first creating with halftoning techniques known in the art such as dithering the parallel and the transverse line halftones, separately for each ink layer. Then the cross-halftone is created by applying an “OR” operation between the parallel and the transverse line halftoned ink layers of the same ink. For cyan, magenta and yellow inks, this results in the three cross-halftoned ink separation layers that are sent to the printer.

Examples of Images Showing Hue Changes, Decolorization or Recolorization and Independent Color Changes

[0151] Let us first schematically show hue changing as well as decolorization or recolorization examples. The butterfly image with hues close to the original hues is shown in FIG. 23A and the hue modified butterfly is shown in FIG. 23B. In the present example, the hues of the background have evolved from green 2301 to violet 2302 and of the wings from red-orange 2305 to blue 2306. The hue modified butterfly appears when rotating under directed light the print in-plane by 90 degrees.

[0152] The “decolorization” example shown in FIGS. 24B and 24A rely on a first original colored image and on a second grayscale image obtained by simply reducing the chroma of the original image colors to zero, i.e. transforming the colored image into a grayscale image. However, for the viewers, it is more spectacular to have the grayscale image as non-rotated image (FIG. 24A) and then by turning the print, let the colors appear (FIG. 24B: recolorization). For recolorization, we can therefore consider the achromatic image as desired non-rotated image (FIG. 24A) and the chromatic image as desired rotated image (FIG. 24B). In FIGS. 24A and 24B, the background remains of the same color 2400, the gray umbrella 2401 become a multi-color umbrella, with magenta 2402, cyan 2403, yellow 2404, orange 2405 and red 2406 sections.

[0153] For security purposes, one may conceive an identity document where the front view of the face appears in color (FIG. 25A, 2501) and, upon rotation becomes gray (FIG. 25B, 2511) and where the side view of the face is gray in non-rotated mode (FIG. 25A, 2502) and upon rotation, disappears and is replaced by text in gray (FIG. 25B, 2512).

[0154] As a first example of independent color changes, FIG. 26A shows two message images embedded into the same cross-halftone print area. A first message image appears in non-rotated mode (FIG. 26B) with background color 2610 and foreground color 2611 and a second message image appears upon 90° in-plane rotation with background color 2620 and foreground color 2621. Since the second message image is laid out in vertical direction (FIG. 26A, 2602), upon 90° in-plane rotation, this second message appears horizontally (FIG. 26C). Background and foreground colors of both message images can be freely chosen. The authenticity of a document incorporating the image with the two messages may be verified by making sure that only the first message appears in non-rotated mode and only the second message appears in rotated mode and that the displayed colors are the ones that should be displayed. In addition, when displaying the non-rotated image and the rotated image in front of a smartphone, the software of the smartphone may recognize the messages and verify by communicating with a server over the Internet that the two messages are authentic.

[0155] A similar example of a print area embedding two independent color images is shown in FIGS. 27A, 27B and 27C, with the goal of showing two independent QR codes of varying colors or gray levels when rotating a same cross-halftone print. Two QR-code images are embedded within one print area by giving as input the two QR code images as either color or grayscale images, to be imaged as independently color varying images. Under directed light, the first QR-code “QR1” is visible in non-rotated mode (FIG. 27B) and the second QR-code “QR2” is visible (FIG. 27C) by applying a 90° in-plane rotation to the print. Let us call the “black” dots of the QR-code image foreground dots and the white dots background dots. In non-rotated mode, the color of the QR1 foreground dots must have a color or grayscale value different from the color of the background dots. In rotated mode, the color of the QR2 foreground dots must have a color or grayscale value different from the color of the background dots. Software running on a smartphone or tablet may acquire the image in rotated and non-rotated modes, apply preprocessing operations to compensate the effect of the projective transformation and to enhance the contrast between foreground and background dots, decode the message content, display it and possibly verify either by interacting with its memory or by communicating with a server over the Internet that the two acquired QR code messages are authentic.

[0156] Let us consider a further example of two independent color pictures embedded within one cross-halftone print image, showing the first picture before rotation and a second independent picture after applying a 90° in-plane rotation to the print. FIG. 28A shows schematically the combined image with the two overlaid independent color pictures, the first picture being shown in the figure in light gray tones and the second picture shown in the figure as dark gray tones. FIG. 28B represents the first color picture, shown in the figure as a gray image and FIG. 28B shows the second picture, shown in the figure as a gray image.

Steps for Producing Color Images that Change their Colors upon In-Plane Rotation

[0157] The steps to be carried out are either part of the preparation phase or part of the printing phase. They can be deduced from the reproduction workflow described in FIG. 22.

[0158] In a preferred embodiment, the steps of the preparation phase are the following:

[0159] a) Establish a 6-ink spectral prediction model for predicting the color printed with cross-halftones with three ink surface coverages (c.sub.P, m.sub.P, y.sub.P) laid out in parallel mode and three ink surface coverages (c.sub.T, m.sub.T, y.sub.T) laid out in transverse mode, i.e. predicting the total reflectance R.sub.total(c.sub.P, m.sub.P, y.sub.P, c.sub.T, m.sub.T, y.sub.T) as a function of the 6 ink surface coverages (c.sub.P, m.sub.P, y.sub.P, c.sub.T, m.sub.T, y.sub.T). The corresponding calculations are described in Section “Spectral prediction model”. The underlying ink-spreading enhanced cellular Yule-Nielsen modified spectral Neugebauer model is adapted to predict the color of line-on-line halftones (FIGS. 10 and 12). The predicted cross-halftone reflectances are converted to colors (e.g. CIELAB) under a certain illuminant, e.g. D65.

[0160] b) Establish a 6D cross-halftone surface coverage to color table by establishing the correspondence between the 6 surface coverages formed by cmy.sub.P=(c.sub.P, m.sub.P, y.sub.P) and cmy.sub.T=(c.sub.T, m.sub.T, y.sub.T) and the non-rotated colors Lab(cmy.sub.P, cmy.sub.T) and rotated colors Lab(cmy.sub.T, cmy.sub.P). This is carried out by first obtaining with the spectral prediction model of step (a) for each considered 6 ink surface coverage the reflectance R.sub.total(c.sub.P, m.sub.P, y.sub.P, c.sub.T, m.sub.T, y.sub.T) and by converting it to a CIELAB color. The 6 ink surface coverages are varied in steps of e.g. 5%.

[0161] c) Establish a 6D desired color table mapping a desired color Lab.sub.inNR before rotation and a desired color Lab.sub.inR after rotation into an optimal set of 6 ink surface coverages, e.g. (c.sub.P, m.sub.P, y.sub.P, c.sub.T, m.sub.T, y.sub.T), using optimization formula (9).

[0162] d) Establish the print gamut and the different constrained color variation sub-gamuts according to Section “Print gamut and constrained color variation sub-gamuts”. The print gamut is obtained by the hull of all colors present in the 6D cross-halftone surface coverage to color table. The other sub-gamuts are obtained as described in that section.

[0163] e) Separately gamut map input colors (e.g. sRGB) covering the color space to the print gamut, to the constrained color variation sub-gamuts, i.e. the hue varying sub-gamut, the decolorization sub-gamut and the independent color variation sub-gamut, by applying a known gamut mapping technique, e.g. the two-foci gamut mapping algorithm [Morovic and Luo 2001]. This yields the print gamut mapping table, the hue-shifting gamut mapping tables, the decolorization gamut mapping table and the independent color variation gamut mapping table. Each of these tables is a “constrained color variation gamut mapping table”.

[0164] The steps that are part of the printing phase are the following:

[0165] f) The input is specified by two desired input images (e.g. sRGB images), a first desired image in non-rotated mode and a second desired input image in rotated mode. Instead of a colored input image in rotated mode, one may also ask for an achromatic rotated image. One may also ask for two grayscale images, or two independent color images, one in-non-rotated and one in rotated mode. Optionally a p value is given to allow for higher chroma values of the obtained non-rotated and rotated hues.

[0166] g) Use the gamut mapping tables obtained in step (e) to perform the gamut mapping of the non-rotated input image colors into the constrained color variation sub-gamuts, i.e. into the hue-shifting sub-gamuts, into the decolorization sub-gamut, or into the independent color variation sub-gamut depending if a hue shifted, an achromatic or an independently colored rotated image is desired and in case of a hue-shifted rotated image, depending on the desired ρ value. This step results in two gamut mapped images, one in non-rotated mode and the second in rotate mode. This gamut mapping step prevents the occurrence of ghosting effects between the non-rotated and the rotated cross-halftone prints.

[0167] h) For each pair of corresponding colors from the two gamut mapped images, deduce with the help the 6D desired color table from step (c) the surface coverages of the 6 inks, e.g. (c.sub.P, m.sub.P, y.sub.P, c.sub.T, m.sub.T, y.sub.T).

[0168] g) From the 6 ink surface coverages, create the cross-halftoned ink separation layers with the surface coverages (c.sub.P, m.sub.P, y.sub.p) in parallel mode indicating the relative width of the vertical ink lines and the surface coverages in transverse mode (c.sub.T, m.sub.T, y.sub.T) mode indicating the relative width of the horizontal ink lines (see FIG. 14A). Send the resulting halftoned ink separation layers to the printer which will print them on a the ink attracting plastic layer laid out on top of a metalized film or of a white diffusing layer.

Computing System for Synthesizing Color Images that Change Colors upon In-Plane Rotation

[0169] A computing system for reproducing color images that change colors upon in-plane rotation comprises a CPU (FIG. 29, 2910), memory (2910), disks (2912) and a network interface (2911) and is operable to run software that comprises callable functions operable to perform the steps described in Section “Steps for producing color images that change their colors upon in-plane rotation” and shown in FIG. 22, i.e. functions which in the preparation phase are operable:

[0170] to create and calibrate a 6-ink cross-halftone spectral prediction model for predicting the color printed with cross-halftones with three ink surface coverages (c.sub.P, m.sub.P, y.sub.P) laid out in parallel mode and three ink surface coverages (c.sub.T, m.sub.T, y.sub.T) laid out in transverse mode

[0171] to establish a 6D cross-halftone surface coverage to color table by expressing the correspondence between the 6 surface coverages formed by cmy.sub.P=(c.sub.P, m.sub.P, y.sub.P) and cmy.sub.T=(c.sub.T, m.sub.T, y.sub.T) and the non-rotated colors Lab(cmy.sub.P, cmy.sub.T) and rotated colors Lab(cmy.sub.T, cmy.sub.P);

[0172] to establish a 6D desired color table mapping a desired color Lab.sub.inNR before rotation and a desired color Lab.sub.inR after rotation into an optimal set of 6 ink surface coverages, e.g. (c.sub.P, m.sub.P, y.sub.P, c.sub.T, m.sub.T, y.sub.T);

[0173] to establish the print gamut, the hue variation sub-gamuts, the decolorization sub-gamut and the independent color variation sub-gamut;

[0174] to separately gamut map input colors (e.g. sRGB) to the print gamut, to the constrained color variation sub-gamuts, i.e. the hue variation sub-gamuts, the decolorization sub-gamut and the independent color variation sub-gamut.

[0175] The functions in the printing phase are operable

[0176] to obtain as input two desired images (e.g. sRGB images), a first desired image in non-rotated mode and a second desired input image in rotated mode;

[0177] to perform, with the help of the gamut mapping tables established in the preparation phase, the gamut mapping of the non-rotated input image colors into the constrained color variation gamuts, i.e. into the hue-shifting sub-gamuts, the decolorization sub-gamut, or into the independent color variation sub-gamut depending if a hue-shifted, an achromatic or an independently colored rotated image is desired and in case of a hue-shifted rotated image, depending on the desired p value;

[0178] for each pair of corresponding colors from the two gamut mapped images, to deduce with the previously established 6D desired color table the surface coverages of the 6 inks, e.g. (c.sub.P, m.sub.P, y.sub.P, c.sub.T, m.sub.T, y.sub.T);

[0179] from the 6 ink surface coverages, to create the cross-halftoned ink separation layers with the surface coverages (c.sub.P, m.sub.P, y.sub.P) in parallel mode indicating the relative thickness of the vertical ink lines and the surface coverages in transverse mode (c.sub.T, m.sub.T, y.sub.T) mode indicating the relative thickness of the horizontal ink lines

[0180] to send the resulting halftoned ink separation layers to the printer which prints them.

Viewing the Cross-Halftone Prints Showing Hue, Chroma and Lightness Variations

[0181] In front of a window or light source, under directed light, a first color or grayscale image appears on the cross-halftone printed with cyan, magenta, yellow inks. Then, upon 90° in plane rotation of the cross-halftone print, either the same image with different colors or a second independent color or grayscale image appears. The succession of these two views upon in-plane rotation of the image identifies the cross-halftone print as authentic. The same succession of two views is obtained upon displacement of the observer, see FIG. 2, when the observer moves from position 212 to position 213 or vice versa. Best results are obtained under directed light, i.e. when light coming from the front is reflected by the cross-halftone print and reaches the eye of the observer. But also under partially diffuse illumination conditions, the succession of color or grayscale images can be observed, possibly with a reduced contrast. In addition, as shown in FIG. 2, it not necessary to observe the print exactly at a 0° or 90° angle in respect to the halftone lines. One may vary the observation angle in respect to the halftone grid by approximately ±25 degrees and still see the corresponding first or second image. FIG. 2 shows a viewing orientation 212 that is not exactly parallel to the transverse halftone lines 214 and perpendicular to the parallel halftone lines 215. Note also that the parallel halftone lines 215 are not exactly parallel to the incident light orientation(s) 211 and the transverse halftone lines 214 are not exactly perpendicular to the incident light orientation(s) 211. Therefore the present invention is well adapted to protect security document that cannot be carefully positioned, for example a parking permit located inside the car, in front of the driver seat, behind the windshield.

Extension of the Presented Methods to Other Ink Colors

[0182] The effects that have been described by relying on cyan, magenta, yellow and white inks can also be realized by cyan, magenta and yellow inks, some of them possibly having zero surface coverages, and additional inks such as light cyan, light magenta, gray, light gray, red, green, blue, orange, violet, pink and black inks. Such an extension may also require to extend the color prediction model in order to predict colors printed with cyan, magenta, yellow and with the additional inks.

Extension of the Presented Methods to a Position-Dependent Small Area Light Source Illuminating the Cross-Halftone Superposed on a Diffusing Substrate

[0183] A cross-halftone printed on a transparent sheet superposed with a diffusing preferably white substrate reflects light in the same incident plane as the incoming light as well as in other directions. It is possible to have a cross-halftone placed horizontally and illuminated by a small area light source or a point light source, such as the flash light of a smartphone. When the small area light source (FIG. 30A, 3000) is in front of the observer and illuminates the cross-halftone print from behind, the first color image that appears to the observer 3003 is the non-rotated image. But when the point light source is placed on the side, for example by rotating it azimuthally by 90 degrees (FIG. 30B, 3010), the second “rotated” color image appears, i.e. the one that would be seen with the light source in front of the observer and the print rotated azimuthally by 90 degrees. In a preferred embodiment, the light source should illuminate the print from a shallow angle, e.g. at 60° in respect to the normal of the print. The observer should view the print at an angle close to the normal of the print (FIG. 31B, 3111), e.g. at a zenith angle of 15° in respect to the normal of the print. At these zenith angles, the observer can view the print from any azimuthal angle and see a substantially same picture.

[0184] The color changing effects can be explained by following the path of light when the point light source (FIG. 31A, 3112) is in front of the print in respect to the observer 3111 and when the ink halftone lines 3101 are parallel to the incident plane. In that case, the observer does not perceive a significant dot gain. Light rays that reach the observer and that enter through the ink areas 3116 also exit through the ink areas 3117, and light rays reaching the observer that enter through the unprinted areas 3118 also exit through the unprinted areas 3119.

[0185] When the light illuminates the print from the right or left side of the observer (FIG. 31B, 3110), for example by rotating the light source azimuthally by 90° or by activating another light source located on the right or left side of the observer, the ink halftone line 3101 is now in transverse mode in respect to the incident light that reaches the observer. There will be a strong dot gain, since a high proportion of light rays enters through the ink 3102, and exits through the paper 3004 or vice-versa. Some of these light rays are reflected towards the vertical or at a narrow zenithal angle in respect to the vertical and reach the observer 3111. Due to the important dot gain, the observer sees the second “rotated” color image upon rotation of the light source.

[0186] In order to illustrate this phenomenon, FIG. 32 shows in a similar manner as in FIG. 5 the color changes that are achieved with a single ink line halftone printed on a transparency of a thickness of 120 microns, superposed on a diffusely reflecting sheet of paper, at an illumination angle of 45° and a measuring angle of 45°, before rotation of the line halftone print, i.e. in parallel mode (empty square) and after in-plane rotation of the line halftone print by 90°, i.e. in transverse mode (filled square). On the same graph, additional measurements are carried out at an illumination angle of 65° and measuring angle of 0° (triangles). The empty triangles or squares indicate the color before rotation of the line halftone (parallel mode) and the filled triangles or squares indicate the color after rotation of the line halftone (transverse mode). When the observer looks at the print from a position close to the normal of the print, rotating the print by 90° or rotating the illuminating light source by 90° produces the same color change.

[0187] Additional measurements of color changes are plotted in FIG. 33 for cross halftones superposed on a white diffusing substrate comprising two inks at 50% nominal surface coverages, one ink halftone forming one set of halftone lines (e.g. parallel lines) and the second ink forming the second set of orthogonal halftone lines (e.g. transverse lines). Further measurements of color changes are plotted in FIG. 34A and FIG. 34B, for cross-halftone lines formed by 3 inks, with two inks forming an ink halftone in one orientation (e.g. parallel) and the third ink forming an ink halftone in the perpendicular orientation (e.g. transverse). FIG. 34A shows the measurements at 45° incident angle and 45° measurement angle and FIG. 34B shows measurements of the same ink halftones at 65° incident angle and 0° measurement angle, i.e. in the normal direction to the print. These measurements show that either by turning the cross-halftone print in-plane by a quarter turn, or possibly by turning the incident light by a quarter turn and observing the print from the top, significant changes in hue, lightness and chroma are observable, close to the color changes observed on a metallic substrate (FIG. 5, FIG. 6 and FIG. 7). Therefore, cross-halftone prints on a transparency superposed with a white diffusing substrate exhibit similar sub-gamuts as the cross-halftone prints formed by a superposition with a metallic substrate. As a result, the methodology presented for cross-halftones superposed with a metallic substrate as well as the method steps mentioned in Section “Steps for producing color images that change their colors upon in-plane rotation” are also valid for cross-halftones printed on a plastic layer superposed with a diffusely reflecting substrate, called diffusely reflecting cross-halftone prints.

[0188] For authenticating security documents and goods, it would be convenient to have a first light source in front of the cross-halftone print superposed with a white diffusely reflecting substrate and a second light source on the right (or left) side of the print, allowing to see by activating the first light source the “non-rotated” color image and by activating the second light source the “rotated” color image.

Embedding Metallic Elements Printed with Transparent Inks, e.g. Cyan, Magenta, Yellow into Security Documents and Valuable Articles

[0189] Cross-halftone prints may be formed by a base (FIG. 4C, 434), a metallic specularly or white diffusely reflecting foil 433, a transparent layer 432 and an ink halftone layer 431 printed with transparent inks, e.g. cyan, magenta, yellow. They can be incorporated into any document e.g. within a small place reserved for special effects on paper banknotes, within a polymer banknote or within the layers of a laminated plastic card. The cross-halftone color prints can also be placed on the data page of a passport.

[0190] An identity card may incorporate the reduced size image of the document holder which changes from color to grayscale depending on the rotation mode, see FIGS. 25A and 25B. The color or grayscale picture of the document holder may be seen in one view (2502), e.g. the non-rotated view and be replaced by text (2512) giving the name and birth date of the document holder in the other view, e.g. the rotated view.

[0191] A cross-halftone image may be printed on a transparent layer affixed on top of a valuable metallic component, such as a car engine, an aerospace metallic component, a component of a weapon, a vehicle component, a machine tooling equipment. An additional transparent protective layer may be pasted or printed on top of the printed cross-halftone. Under directed light, the observed cross-halftone printed metallic piece will then show a first image before and a second image after 90° in-plane rotation of the printed piece in respect to the observer.

[0192] In case of a cross-halftone superposed with a diffusing layer, the cross-halftone print may remain fixed. Viewed under a light source in front of the print and of the observer (e.g. FIG. 30A), the first “non-rotated” image appears. Under a light source positioned on the side of the observer (e.g. FIG. 30B), the second “rotated” image appears.

[0193] The cross-halftone prints images may incorporate hard to reproduce information, such as different superposed pieces of text as first “non-rotated” and second “rotated” image or different superposed codes, such as QR-codes as first “non-rotated” and second “rotated” images. Such prints on are very difficult to reproduce without the software that is able to generate cross-halftone images. With the software enabling carrying out the steps mentioned in Section “Steps for producing color images that change their colors upon in-plane rotation” and illustrated in FIG. 22, potential counterfeiter will not be able to avoid ghosting effects, i.e. the simultaneous mixed presence of elements of the first “non-rotated” and the second “rotated” image. These ghosting effects will indicate that the faked cross-halftone image is not genuine.

[0194] The cross-halftone print providing the means to authenticate a valuable article can be part of its package or can be part of the label describing the valuable article. For example, a package containing drugs may use as a base the metallic cross-halftone print, coated with a protecting transparent layer. The metallic cross-halftone print may incorporate an alphanumeric code, a barcode or a QR-code that disappears and appears, or that changes from a first code to a second code when rotating the print in-plane by 90°. This code may be used as a further authentication means, by capturing the photograph of the print with a camera hooked onto a computing device such as a smartphone, connecting it to a Web server and verifying that the code (alphanumeric, barcode or QR-code) is valid. Another possibility is to have in the cross-halftone print the message “ORIGINAL DRUG” hidden in non-rotated mode and appearing in rotated mode or vice-versa. The message may also be hidden in non-rotated mode by another message (FIGS. 26A, 26B, 26C) or within a color picture.

Benefits of the Present Invention

[0195] In contrast to many other counterfeit prevention features, cross-halftone prints with hue changing, color to grayscale, grayscale to color varying images or images switching from a first to a second independent image upon quarter turn in plane rotation can be authenticated by the naked eye, without requiring additional devices, such as a smartphone, a camera, a scanner or a UV illumination. Similar color changing effects are observable upon switching between a light source illuminating the cross-halftone prints from the front and a light source illuminating the cross-halftone prints from the side.

[0196] These cross-halftone prints cannot be counterfeited without the software optimizing all ink surface coverages in parallel and transverse mode, e.g. the 6 parallel and transverse cyan (c), magenta (m), yellow (y), ink surface coverages. Without this optimization, desired color changes upon in-plane rotation of the print or upon change of illumination orientation cannot occur and/or strong ghosting effects appear. Ghosting has the effect of showing the first and the second image mixed in both the non-rotated and rotated viewing modes.

[0197] If a doubt arises about the authenticity of a document or valuable item incorporating the cross-halftone print with the color varying images, one may examine the cross-halftone print under a microscope and verify that the cross-halftones (FIG. 3, 302) are indeed present, i.e. verify the presence of transverse and parallel line halftones in respect to the incident plane.

[0198] Cross-halftone color prints exhibiting a color changing image or showing a first image switching to a second independent image create surprising effects that can be used in amusement parks, for decoration, for art or for advertisements. For example, a large color metallic poster placed at a position where it is illuminated from one side will show to a person walking by (the “observer”) first a non-rotated view and then a rotated view (FIG. 2, 212 and 213, respectively). The metallic poster then displays in the first view one image and in the second view a second independent image. Or it may show in the first view a color image where the message is hidden and in the second view the same image, but with the hidden message revealed (FIGS. 28A and 28B, right side). Such effects are induced by the change of position of the observer. This unexpected succession of views raises the attention of the observer.

[0199] Cross-halftone prints with images that vary their colors upon change of the observation conditions have a high decorative value and may be used for the decoration of house walls, museum walls, exhibitions, and of goods having a high symbolic value such as cars, watches, jewelry, smartphones, tablets, portable computers, fashion articles, toys, expensive drinks and cosmetic articles.