Device for acquiring bimodal images

Abstract

The invention relates to an image acquisition device comprising a sensor composed of an array of photosensitive pixels, and an array of elementary filters covering the sensor. Pixels may be of three different types: panchromatic pixels, primary colour pixels and infrared pixels. Under low illumination conditions, the device displays a monochrome image using panchromatic pixels, and under high illumination conditions, a colour image with a high signal to noise ratio, by combining primary colour images and subtracting the infrared image.

Claims

1. Image acquisition device comprising a sensor composed of an array of photosensitive elements called pixels, an array of elementary filters covering said sensor, each elementary filter being associated with a corresponding pixel, and a processor, wherein: the elementary filter array comprises primary colour filters, where a primary colour filter only transmits one primary colour, filters transmitting the entire visible spectrum called panchromatic filters, and infrared filters transmitting the near infrared, the panchromatic filters representing at least 50% of said elementary filters; the processor configured to: calculate an average luminance on a set of panchromatic pixels to determine if at least one sensor zone is under low or high illumination conditions; if said zone is under low illumination conditions, form a monochrome image of said zone using the panchromatic pixels; if said zone is under high illumination conditions, form a colour image of said zone using primary colour pixels and subtract an infrared image from said zone obtained from infrared pixels.

2. Image acquisition device according to claim 1, wherein the monochrome image of said zone is obtained by interpolation of panchromatic pixels forming part of this zone.

3. Image acquisition device according to claim 1, wherein the colour image of said zone is obtained by combination of primary colour images of said zone, a primary colour image being obtained by interpolation of primary colour pixels of said zone.

4. Image acquisition device according to claim 1, wherein the infrared image of said zone is obtained by interpolation of infrared pixels of said zone.

5. Image acquisition device according to claim 1, wherein panchromatic filters represent at least 75% of the elementary filters.

6. Image acquisition device according to claim 5, wherein the elementary filter array is generated by two-dimensional periodic repetition of the following pattern: $F_{d\&n}=\left(\begin{array}{cccc}R& W& G& W\\ W& W& W& W\\ \mathrm{IR}& W& B& W\\ W& W& W& W\end{array}\right)$ where R,G,B represent transmission filters in the red, the green and blue respectively, IR is an infrared filter and W is a panchromatic filter, the pattern being defined within one permutation of R,G,B,IR.

7. Image acquisition device according to claim 5, wherein the elementary filter array is generated by two-dimensional periodic repetition of the following pattern: $F_{d\&n}=\left(\begin{array}{cccc}\mathrm{Ye}& W& \mathrm{Ma}& W\\ W& W& W& W\\ \mathrm{IR}& W& \mathrm{Cy}& W\\ W& W& W& W\end{array}\right)$ where Ye, Ma, Cy represent transmission filters in yellow, magenta and cyan respectively, IR is an infrared filter and W is a panchromatic filter, the pattern being defined within one permutation of Ye, Ma, Cy, IR.

8. Image acquisition device according to claim 1, wherein the elementary filter array is generated by a two-dimensional periodic repetition of the following pattern: $F_{d\&n}=\left(\begin{array}{cccc}W& R& W& G\\ R& W& \mathrm{IR}& W\\ W& G& W& B\\ \mathrm{IR}& W& B& W\end{array}\right)$ where R,G,B represent transmission filters in the red, green and blue respectively, IR is an infrared filter and W is a panchromatic filter, the pattern being defined within one permutation of R,G,B,IR.

9. Image acquisition device according to claim 1, wherein the elementary filter array is generated by a two-dimensional periodic repetition of the following pattern: $F_{d\&n}=\left(\begin{array}{cccc}W& \mathrm{Ye}& W& \mathrm{Ma}\\ \mathrm{Ye}& W& \mathrm{IR}& W\\ W& \mathrm{Ma}& W& \mathrm{Cy}\\ \mathrm{IR}& W& \mathrm{Cy}& W\end{array}\right)$ where Ye, Ma, Cy represent transmission filters in the yellow, magenta and cyan respectively, IR is an infrared filter and W is a panchromatic filter, the pattern being defined within one permutation of Ye, Ma, Cy, IR.

10. Image acquisition device according to claim 1, wherein the processor determines that said zone has low illumination if the average luminance on all panchromatic pixels in said zone is less than a first threshold and said zone has high illumination if the average luminance on all panchromatic pixels in said zone is higher than a second threshold, the second threshold being higher than the first threshold.

11. Image acquisition device according to claim 10, wherein if the average of the luminance on all panchromatic pixels in said zone is between the first and second thresholds, the processor makes a combination of the monochrome image and the colour image of said zone.

12. Image acquisition device according to claim 1, wherein the sensor zone is the entire sensor and the processor determines if the sensor is under low illumination or high illumination conditions.

13. Image acquisition device according to claim 1, wherein the monochrome image of the sensor is segmented into homogeneous regions and the processor operates independently on each homogeneous zone of the sensor.

14. Image acquisition device according to claim 1, wherein when a zone is under low illumination conditions, a test is performed to determine if the intensity of infrared pixels forming part of this zone is greater than a predetermined IR threshold, and if so, Infrared pixels represented in false colours are superposed with the monochrome image of this zone.

15. Image acquisition device according to claim 1, wherein when a zone is under low illumination conditions, sub-zones with homogeneous intensity in the infrared spectrum are identified in this zone, and for each sub-zone thus identified, it is determined if the average intensity of infrared in this sub-zone is greater than a predetermined IR threshold, and if so, this sub-zone is represented by a false colour image based on infrared pixels in this sub-zone, the false colour image of said sub-zone then being represented superposed with the monochrome image of the zone of which if forms part.

16. Image acquisition device according to claim 1, wherein the false colour image of said sub-zone is obtained by interpolation of the infrared pixels forming part of this sub-zone.

17. Image acquisition device according to claim 1, wherein the infrared filters transmit the 700nm-1100nm band.

18. Image acquisition device according to claim 1, wherein the infrared filters transmit the 700nm-1700nm band.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) Other characteristics and advantages of the invention will become clear after reading the description of preferred embodiments of the invention with reference with the attached figures among which:

(2) FIG. 1 diagrammatically shows a first colour filter array known in the state of the art;

(3) FIG. 2 diagrammatically shows a second colour filter array known in the state of the art;

(4) FIG. 3A diagrammatically shows an example filter array for an image acquisition device according to a first example embodiment of the invention;

(5) FIG. 3B diagrammatically shows an example filter array for an image acquisition device according to a second example embodiment of the invention;

(6) La FIG. 4A diagrammatically shows an example filter array for an image acquisition device according to a third example embodiment of the invention;

(7) La FIG. 4B diagrammatically shows an example filter array for an image acquisition device according to a fourth example embodiment of the invention;

(8) FIG. 5 diagrammatically shows processing of pixels for an image acquisition device according to a first embodiment of the invention;

(9) FIG. 6 diagrammatically shows processing of pixels for an image acquisition device according to a second embodiment of the invention.

(10) FIG. 7 diagrammatically shows an image acquisition device according to the present invention.

DETAILED PRESENTATION OF PARTICULAR EMBODIMENTS

(11) We will once again consider an image acquisition device in the following comprising a sensor composed of an array of elementary photosensitive sensors (CCD or CMOS) also called pixels, and an elementary filter array covering the sensor, each elementary filter being associated with a corresponding pixel.

(12) Such an image acquisition device is illustrated in FIG. 7. It shows the sensor, 710, composed of an array of elementary photosensitive sensors, an array of elementary filters, 720, covering said sensor, each elementary filter being associated with a corresponding pixel, and a processor, 730.

(13) As detailed below, the array of elementary filters comprises primary colour filters, where a primary colour filter only transmits one primary colour, filters transmitting the entire visible spectrum called panchromatic filters, and infrared filters transmitting the near infrared, the panchromatic filters representing at least 50% of said elementary filters.

(14) Furthermore, the processor is configured to: calculate an average luminance on a set of panchromatic pixels to determine if at least one sensor zone is under low or high illumination conditions; if said zone is under low illumination conditions, form a monochrome image of said zone using the panchromatic pixels; if said zone is under high illumination conditions, form a colour image of said zone using primary colour pixels and subtract an infrared image from said zone obtained from infrared pixels.

(15) Elementary filters in the array are transmission filters of three different types: colour filters, transparent (or panchromatic) filters and infrared filters.

(16) Colour elementary filters allow a set of primary colours to pass through them enabling additive or subtractive synthesis of all colours in the visible spectrum. Thus colour filters may be R,G,B filters as defined in the introduction in the case of an additive synthesis, or Ye,Ma,Cy filters where Ye allows yellow to pass, Ma allows magenta to pass, and Cy allows cyan to pass, in the case of a subtractive synthesis. Other primary colour bases could be envisaged by those skilled in the art without going outside the scope of this invention.

(17) Panchromatic elementary filters allow the entire visible spectrum to pass. In practice they may be made by simple stencils in the array or transparent filters (in the visible). In this case pixels located under these panchromatic elementary filters receive unfiltered light.

(18) Infrared elementary filters only allow a spectral band in the near infrared to pass, for example in the [700 nm-1700 nm] band, and more particularly in the [700 nm-1100 nm] band.

(19) Elementary filters are advantageously arranged in the form of a periodically repeating pattern along two distinct directions (usually orthogonal) in the plane of the sensor.

(20) The elementary filters may be made from a polymer material or may be interference filters in a manner known in itself.

(21) Preferably, the proportion of panchromatic elementary filters in the array will be greater than or equal to 50%, the other elementary filters (colour filters and infrared filters) being distributed in equal proportions. Advantageously, the proportion of panchromatic elementary filters will be equal to 75% and the proportion of each of the other elementary filters will be 6.25%.

(22) FIG. 3A shows an array of elementary filters for an image acquisition system according to a first example embodiment of the invention. There is one sensor pixel for each elementary filter.

(23) In this case, the colour elementary filters are R,G,B filters. Infrared elementary filters are denoted IR and panchromatic filters are denoted W .

(24) It will be noted that the elementary array is generated by a two-dimensional periodic repetition of the basic 4×4 pattern:

(25) $\begin{array}{cc}{F}_{d\&n}=\left(\begin{array}{cccc}R& W& G& W\\ W& W& W& W\\ \mathrm{IR}& W& B& W\\ W& W& W& W\end{array}\right)& \left(3\right)\end{array}$

(26) Variants of this array can be obtained by permutation of the R,G,B,IR filters in the pattern (3).

(27) FIG. 3B shows an array of elementary filters for an image acquisition system according to a second example embodiment of the invention.

(28) In this case, the colour elementary filters are the Ye,Ma,Cy filters and the basic pattern is then:

(29) $\begin{array}{cc}{F}_{d\&n}=\left(\begin{array}{cccc}\mathrm{Ye}& W& \mathrm{Ma}& W\\ W& W& W& W\\ \mathrm{IR}& W& \mathrm{Cy}& W\\ W& W& W& W\end{array}\right)& \left(4\right)\end{array}$

(30) As in the first example, variants of this array may be obtained by permutation of the Ye,Ma,Cy,IR filters in the array (4).

(31) In the examples in FIGS. 3A and 3B, it will be noted that the proportion of panchromatic filters in the array is 75% and the proportion of each colour and infrared filter is 6.25%.

(32) FIG. 4A shows an elementary filter array for an image acquisition system according to a third example embodiment of the invention.

(33) The elementary filters used in the array are R,G,B,IR filters, the basic pattern in this case being:

(34) $\begin{array}{cc}{F}_{d\&n}=\left(\begin{array}{cccc}W& R& W& G\\ R& W& \mathrm{IR}& W\\ W& G& W& B\\ \mathrm{IR}& W& B& W\end{array}\right)& \left(5\right)\end{array}$

(35) As in the first example, variants of this array may be obtained by permutation of the R,G,B,IR filters in the pattern (5).

(36) FIG. 4B shows an elementary filter array for an image acquisition system according to a fourth example embodiment of the invention.

(37) The elementary filters making up the array are the Ye,Ma,Cy,IR filters and the basic pattern is:

(38) 0$\begin{array}{cc}{F}_{d\&n}=\left(\begin{array}{cccc}W& \mathrm{Ye}& W& \mathrm{Ma}\\ \mathrm{Ye}& W& \mathrm{IR}& W\\ W& \mathrm{Ma}& W& \mathrm{Cy}\\ \mathrm{IR}& W& \mathrm{Cy}& W\end{array}\right)& \left(6\right)\end{array}$
As in the second example, variants of this array can be obtained by permutation of Ye,Ma,Cy,IR filters in the pattern (6).

(39) In the examples in FIGS. 4A and 4B, it will be noted that the proportion of panchromatic filters in the array is only 50% and the proportion of each colour and infrared filter is 12.5%.

(40) FIG. 5 diagrammatically shows processing of pixels in an image acquisition system according to a first embodiment of the invention.

(41) In the following, an operation on the pixels will be understood as an operation on the signals received by these pixels.

(42) In a first step 510, the average luminance Ī of the image acquired by the device is estimated by calculating the average of the panchromatic pixels.

(43) In 520, the average thus obtained is compared with a predetermined threshold value, I.sub.th. The acquisition device operates in day (or high illumination) mode if the average exceeds said threshold value, otherwise it operates in night (low illumination) mode.

(44) If the acquisition device is operating in day mode, colour pixels are interpolated in 530 so as to obtain one image for each primary colour, R,G,B or Ye,Ma,Cy. More precisely, as seen above, since the proportion of pixels of a given colour is relatively small, the value of the missing pixels is obtained by interpolation of nearby pixels present in the array. For example, with reference to FIG. 3A, the blue pixel 310 is missing (because it is located under a red elementary filter), and is therefore obtained by interpolation between the four adjacent blue pixels 311, 312, 313, 314.

(45) Primary colour images may optionally be processed for image sharpening. For example, a monochrome image can be obtained by interpolating panchromatic pixels and this monochrome image can be combined with each primary colour image, possibly after high pass filtering. The proportion of panchromatic pixels in the array is higher than the proportion of colour pixels, therefore the resolution of the primary colour images is improved.

(46) In step 540, an image in the near infrared is obtained by interpolation between infrared pixels in the array using the same principle as described above for colour pixels.

(47) In step 550, the primary colour images are added (for an additive synthesis) to obtain a colour image and the infrared image is subtracted from it pixel by pixel. More generally, primary colour images are combined to obtain a colour image. Subtraction of the infrared image can considerably improve the signal to noise ratio. If necessary, the infrared image weighted by weighting factors (similar or identical) is subtracted from each primary colour image to obtain noise free primary colour images before summating them (or more generally combining them), so as to prevent saturation problems. The resulting image can then be displayed or stored in a memory for subsequent processing.

(48) Step 560 corresponds to operation in night mode (or under low illumination). In this mode, a monochrome image is obtained by interpolation of panchromatic pixels. Since there is a majority of panchromatic pixels in the array, a high degree of resolution is obtained. This image can be displayed or stored in a memory for subsequent processing.

(49) In all cases, processing terminates at 570.

(50) According to the first embodiment, processor 710 of the image acquisition device is configured to carry out processing steps 510-570 described above.

(51) It can be understood that the processing of pixels described above is very different if the image acquisition device is operating in day mode or in night mode, and switching from one mode to the other is made by comparison of the average luminance with a predetermined threshold.

(52) According to one first variant of this embodiment, switching from one mode to the other takes place with hysteresis to avoid switching noise (chattering). To achieve this, a first luminance threshold is set for the transition from day mode to night mode and a second luminance threshold is set for the transition from night mode to day mode, the first threshold being chosen to be less than the second threshold.

(53) According to a second variant of the first embodiment, switching from one mode to the other takes place progressively passing through a transition phase. Thus, the image acquisition device operates in night mode when the average luminance is less than a first threshold and in day mode when it is greater than a second threshold, the second threshold being chosen to be higher than the first threshold. When the average luminance is between the first and the second thresholds, the image acquisition device performs a linear combination of the image obtained by processing in day mode and that obtained by processing in night mode, the weighting coefficients being given by the differences in average luminance with the first and second thresholds respectively.

(54) FIG. 6 diagrammatically shows processing of pixels in an image acquisition device according to a second embodiment of the invention.

(55) Unlike the first embodiment, switching between day mode (or high illumination) and night mode (low illumination) is active on homogeneous regions of an acquired image, each homogeneous region being processed separately.

(56) In step 603, a monochrome image is obtained by interpolating panchromatic pixels. More precisely, for each missing panchromatic pixel in the array (for example colour or infrared pixel), an intensity is calculated from nearby panchromatic pixels.

(57) In step 605, the monochrome image thus obtained is segmented into homogeneous regions. This can be done using a switching technique known in the state of the art, for example like that described in the article by S. Tripathi et al. entitled “Image Segmentation: a review” published in International Journal of Computer Science and Management Research, vol. 1, No. 4, November 2012, pp. 838-843. The result obtained is thus P homogeneous regions.

(58) In step 610, the average luminance denoted Ī.sub.p is calculated for each region Z.sub.p, p=1, . . . , P.

(59) In step 620, the averages of the luminance thus obtained are compared with a predetermined threshold value. If the average Ī.sub.p exceeds said threshold value, the region Z.sub.p is considered to be high illumination. Conversely, if the average Ī.sub.p is less than said threshold value, the region Z.sub.p is considered to be low illumination.

(60) Note that for a given zone, the distinction between low illumination and high illumination depends on the average of the luminance as obtained from panchromatic pixels. In other words, the average luminance is obtained on the entire spectrum.

(61) In step 630, regions with high illumination are processed. For each high illumination region, colour pixels belonging to this region are interpolated so as to obtain an image for this region for each primary colour R,G,B or Ye,Ma,Cy.

(62) In step 640, an image in the near infrared of each high illumination region is obtained by interpolating infrared pixels belonging to this region.

(63) In step 650, primary colour images are combined for each high illumination region (simple summation in the case of an additive synthesis) and the infrared image is subtracted pixel by pixel. The result obtained is thus colour pixels forming a colour image for each high illumination region.

(64) In step 660, the panchromatic pixels of high illumination regions in the monochrome image obtained in step 603 are replaced by colour pixels obtained in step 650. Alternately, a linear combination of the panchromatic pixel (in other words the pixel in the monochrome image) and the colour pixel obtained in step 650 is made for each pixel belong to a high illumination region. Thus, for low illumination regions, the monochrome image (obtained in step 603) is superposed with the colour image (obtained in 650).

(65) According to the second embodiment, processor 710 of the image acquisition device is configured to carry out processing steps 603, 605, 610-660 described above.

(66) As in the first embodiment, in a first variant, two distinct thresholds can be used, a first threshold to detect when a region changes from a low illumination level to a high illumination level, and a second threshold to determine when a region changes from a high illumination level to a low illumination level, the first threshold being less than the second threshold.

(67) It would also be possible in a second variant to use a transition phase between the two thresholds. In other words, below the first threshold the region is represented in monochrome mode by interpolation of panchromatic pixels, and above the second threshold the region is represented in colour, and between the two thresholds the region may be represented by a linear combination of the monochrome image and the colour image.

(68) It can thus be understood that there is an image acquisition device capable of functioning independently of illumination conditions while having a very good signal to noise ratio.

(69) According to another variant not shown of the second embodiment, when test 620 has determined that a zone has low illumination, it is determined if the intensity of infrared pixels forming part of this zone is greater than a predetermined IR threshold. Infrared pixels for which the intensity is greater than the IR threshold are then represented in false colours superposed with the monochrome image of the zone. Thus, if a sub-zone of the low illumination zone has high intensity in the infrared domain, it can be easily identified in the monochrome image of the zone in question. Also alternately, the criterion for switching between a monochrome display and a display in false colours can be made by homogenous sub-zones of the low illumination zone. In this case, the low illumination zone is firstly segmented based on infrared pixels, after interpolation, using the same principle that was described with relation to steps 603 and 605. For each homogeneous sub-zone as defined above, the average intensity of infrared pixels in this sub-zone is calculated and is compared with the above-mentioned IR threshold. When this average is greater than the IR threshold, the sub-zone is represented with false colours superposed on the monochrome image of the zone. The image of the sub-zone in false colours is obtained by interpolation of infrared pixels in this sub-zone. It will thus be understood that in this case, the image of the low illumination zone but with sub-zones with high infrared intensity, is obtained by a combination of the panchromatic pixels (monochrome part) and infrared pixels (part in false colours) excluding elementary colour pixels (RGB) or (Cyan, Magenta, Yellow).