Image decoding apparatus and method for handling intra-image predictive decoding with various color spaces and color signal resolutions

Abstract

The present invention is directed to an image information decoding apparatus adapted for performing infra-image decoding based on resolution of color components and color space of an input image signal. An intra prediction unit serves to adaptively change block size in generating a prediction image based on a chroma format signal indicating whether resolution of color components is one of 4:2:0 format, 4:2:2 format, and 4:4:4 format, and a color space signal indicating whether color space is one of YCbCr, RGB, and XYZ. An inverse orthogonal transform unit and an inverse quantization unit serve to also change orthogonal transform technique and quantization technique in accordance with the chroma format signal and the color space signal. A decoding unit decodes the chroma format signal and the color space signal to generate a prediction image corresponding to the chroma format signal and the color space signal.

Claims

1. A decoding apparatus for decoding a bitstream comprising an encoded image signal that corresponds to an input image signal, the input image signal including a color difference signal and being of a 4:2:2 format and in a YCbCr color space, the apparatus comprising: circuitry configured to decode the encoded image signal to generate quantized transform coefficients; perform inverse-quantization of the decoded quantized transform coefficients to generate 2×4 chroma DC blocks of transform coefficients constituted by collecting only chroma DC coefficients of eight 4×4 transform coefficient blocks within 8×16 pixel blocks with 8×8 pixel blocks aligned in a vertical direction, perform an inverse transform process that acts on the inversely quantized transform coefficients by performing an inverse orthogonal transform in 2×4 block units on the 2×4 chroma DC blocks, and performing a further inverse orthogonal transform, with the inversely orthogonally transformed 2×4 chroma DC block coefficients as the DC coefficients, of each of the eight 4×4 blocks within the 8×16 pixel blocks, and generating a prediction image using an output signal from the inverse transform process, the prediction image including a prediction color difference signal composed in 8×16 pixel block units; wherein the inverse-quantization is performed using a scale in inverse quantization whose value corresponds to a scale of the further inverse orthogonal transform.

2. A decoding method for decoding a bitstream comprising an encoded image signal that corresponds to an input image signal, the input image signal including a color difference signal and being of a 4:2:2 format and in a YCbCr color space, the method comprising: a decoding process comprising decoding the encoded image signal to generate quantized transform coefficients; an inverse quantizing process comprising performing inverse-quantization of the quantized transform coefficients decoded in the decoding process to generate 2×4 chroma DC blocks of transform coefficients constituted by collecting only chroma DC coefficients of eight 4×4 transform coefficient blocks within 8×16 pixel blocks with 8×8 pixel blocks aligned in a vertical direction; an inverse transform process that acts on the transform coefficients inversely quantized by the inverse quantizing process, the inverse transform process comprising: performing an inverse orthogonal transform in 2×4 block units on the 2×4 chroma DC blocks; and performing a further inverse orthogonal transform, with the inversely orthogonally transformed 2×4 chroma DC block coefficients as the DC coefficients, of each of the eight 4×4 blocks within the 8×16 pixel blocks; and an intra-image prediction process comprising generating a prediction image using an output signal of the inverse transform process, the prediction image including a prediction color difference signal composed in 8×16 pixel block units, wherein the inverse quantizing process uses a scale in inverse quantization whose value corresponds to a scale of the further inverse orthogonal transform.

3. A non-transitory computer-readable medium having embodied thereon a program, which when executed by a computer causes the computer to execute a decoding method for decoding a bitstream comprising an encoded image signal that corresponds to an input image signal, the input image signal including a color difference signal and being of a 4:2:2 format and in a YCbCr color space, the method comprising: a decoding process comprising decoding the encoded image signal to generate quantized transform coefficients; an inverse quantizing process comprising performing inverse-quantization of the quantized transform coefficients decoded in the decoding process to generate 2×4 chroma DC blocks of transform coefficients constituted by collecting only chroma DC coefficients of eight 4×4 transform coefficient blocks within 8×16 pixel blocks with 8×8 pixel blocks aligned in a vertical direction; an inverse transform process that acts on the transform coefficients inversely quantized by the inverse quantizing process, the inverse transform process comprising: performing an inverse orthogonal transform in 2×4 block units on the 2×4 chroma DC blocks; and performing a further inverse orthogonal transform, with the inversely orthogonally transformed 2×4 chroma DC block coefficients as the DC coefficients, of each of the eight 4×4 blocks within the 8×16 pixel blocks; and an intra-image prediction process comprising generating a prediction image using an output signal of the inverse transform process, the prediction image including a prediction color difference signal composed in 8×16 pixel block units, wherein the inverse quantizing process uses a scale in inverse quantization whose value corresponds to a scale of the further inverse orthogonal transform.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) FIG. 1 is a block diagram showing outline of the configuration of a conventional image information encoding apparatus adapted for realizing image compression by orthogonal transform such as discrete cosine transform or Karlnen-Loeve transform, etc. and motion prediction/compensation.

(2) FIG. 2 is a block diagram showing outline of the configuration of a conventional image information decoding apparatus corresponding to the above-mentioned image information encoding apparatus.

(3) FIG. 3 is a view for explaining four intra-prediction modes in JVT Codec.

(4) FIG. 4 is a view showing the state where DC coefficients of four 4×4 blocks within 8×8 block are collected to constitute 2×2 block.

(5) FIG. 5 is a block diagram showing outline of the configuration of an image information encoding apparatus according to the present invention.

(6) FIG. 6 is a block diagram showing one example of the configuration of intra-prediction unit in the image information encoding apparatus according to the present invention.

(7) FIG. 7 is a view showing one example of the configuration of orthogonal transform unit in the image information encoding apparatus according to the present invention.

(8) FIG. 8 is a view showing the state where DC coefficients of eight 4×4 blocks within two 8×8 blocks successive in a longitudinal direction are collected to constitute 2×4 blocks.

(9) FIG. 9 is a block diagram showing one example of the configuration of quantization unit in the image information encoding apparatus according to the present invention.

(10) FIG. 10 is a block diagram showing one example of the configuration of inverse-quantization unit in the image information encoding apparatus according to the present invention.

(11) FIG. 11 is a block diagram showing one example of the configuration of inverse-orthogonal transform unit in the image information encoding apparatus according to the present invention.

(12) FIG. 12 is a block diagram showing outline of the configuration of an image information decoding apparatus according to the present invention.

BEST MODE FOR CARRYING OUT THE INVENTION

(13) While practical embodiments to which the present invention is applied will now be described in detail with reference to the attached drawings, it should be noted that the present invention not limited to such embodiments, but it is a matter of course that various changes or modifications can be made within the scope which does not depart from the gist of the present invention.

(14) (1) Configuration and Operation of the Image Information Encoding Apparatus

(15) First, outline of the configuration of the image information encoding apparatus according to the present invention is shown in FIG. 5. The image information encoding apparatus 10 comprises, as shown in FIG. 5, an A/D (Analogue/Digital) converting unit 11, an image sorting buffer 12, an adder 13, an orthogonal transform unit 14, a quantization unit 15, a reversible encoding unit 16, a storage buffer 17, an inverse quantization unit 18, an inverse orthogonal transform unit 19, an adder 20, a frame memory 21, a motion prediction/compensation unit 22, an intra-prediction unit 23, and a rate control unit 24.

(16) In FIG. 5, the A/D converting unit 11 converts an inputted image signal into a digital signal. Further, the image sorting buffer 12 performs sorting of frames in accordance with GOP (Group of Pictures) structure of image compressed information outputted from the image information encoding apparatus 10. In this example, the image sorting buffer 12 delivers image information of the entirety of frames to the orthogonal transform unit 14 in regard to images in which intra (intra-image) encoding is performed. The orthogonal transform unit 14 implements orthogonal transform such as discrete cosine transform or Karhunen-Loeve transform, etc. to the image information to deliver transform coefficients to the quantization unit 15. The quantization unit 15 implements quantization processing to the transform coefficients delivered from the orthogonal transform unit 14.

(17) The reversible encoding unit 16 implements reversible encoding such as variable length encoding or arithmetic encoding, etc. to the quantized transform coefficients to deliver the transform coefficients thus encoded to the storage buffer 17 to store them thereinto. The encoded transform coefficients are outputted as image compressed information.

(18) The behavior (operation) of the quantization unit 15 is controlled by the rate control unit 24. Moreover, the quantization unit 15 delivers quantized transform coefficients to the inverse quantization unit 18. The inverse quantization unit 18 inverse-quantizes the transform coefficients thus delivered. The inverse orthogonal transform unit 19 implements inverse orthogonal transform processing to the inverse-quantized transform coefficients to generate decoded image information to deliver the information thus generated to the frame memory 21 to store them thereinto.

(19) On the other hand, the image sorting buffer 12 delivers image information to the motion prediction/compensation unit 22 in regard to images in which inter (inter-image) encoding is performed. The motion prediction/compensation unit 22 takes out, from the frame memory 21, image information referred at the same time to implement motion prediction/compensation processing thereto to generate reference image information. The motion prediction/compensation unit 22 delivers the reference image information thus generated to the adder 13. The adder 13 converts the reference image information into a difference signal between the reference image information and corresponding image information. In addition, the motion compensation/prediction unit 22 delivers motion vector information to the reversible encoding unit 16 at the same time.

(20) The reversible encoding unit 16 implements reversible encoding processing such as variable length encoding or arithmetic encoding, etc. to the motion vector information thus delivered to form information inserted into header portion of image compressed information. It is to be noted that since other processing are the same as those of image compressed information to which intra-encoding is implemented, the explanation thereof will be omitted.

(21) In this example, in the above-described JVT Codec, in performing intra-encoding, there is employed intra-predictive encoding system of generating prediction images from pixels around block to encode differences therebetween. Namely, in regard to images in which intra-encoding is performed (I picture, I slice, intra macro block, etc.), prediction image is generated from already encoded pixel values in the vicinity of pixel block to be encoded so that difference with respect to the prediction image is encoded. The inverse quantization unit 18 and the inverse orthogonal transform unit 19 respectively inverse-quantize and inverse orthogonally transform the intra-encoded pixels. The adder 20 adds output of the inverse orthogonal transform unit 19 and prediction image used in encoding corresponding pixel block to deliver added value thus obtained to the frame memory 21 to store it thereinto. In the case of pixel block to be intra-encoded, the intra prediction unit 23 reads out already encoded neighboring pixels stored in the frame memory 21 to generate prediction image. At this time, also with respect to intra prediction mode used in generation of prediction image, reversible encoding processing is implemented thereto at the reversible encoding unit 16 to provide an output in the state included in image compressed information.

(22) (2) The Part to Which the Present Invention is Applied in the Image Information Encoding Apparatus

(23) (2-1) Intra Prediction Unit

(24) An example of the configuration of the intra prediction unit 23 is shown in FIG. 6. The intra prediction unit 23 switches prediction technique on the basis of chroma format signal indicating whether resolution of color component is that of any one of 4:2:0 format, 4:2:2 format and 4:4:4 format, etc., and color space signal indicating whether color space is any one of YCbCr, RGB and XYZ, etc. In this example, the chroma format signal and the color space signal are set in advance by external user, etc., and are delivered to the image information encoding apparatus 10.

(25) In the intra prediction unit 23 shown in FIG. 6, the chroma format signal and the color space signal are delivered to switches 30, 32. The switches 30 and 32 select any one of intra predictors 31a, 31b, 31c on the basis of the chroma format signal and the color space signal to deliver an image signal which has been read out from the frame memory 21 to the selected intra predictor to output prediction image from the selected intra predictor. The switches 30, 32 select the same intra predictor. It is to be noted that while explanation has been given in FIG. 6 on the premise that any one of three kinds of intra predictors 31a, 31b, 31c is selected, the number of intra predictors, i.e., the number of prediction systems may be arbitrarily set.

(26) (2-1-1)

(27) First, the operation of the intra predictor 31a will be explained. The intra predictor 31a serves to perform prediction with 8×8 block being as unit with respect to an image signal in which the chroma format signal indicates 4:2:0 format and color space signal indicates YCbCr. It is to be noted that since the operation of the intra predictor 31a is the same as that of the previously described prior art, the detailed explanation thereof is omitted.

(28) (2-1-2)

(29) Then, the operation of the intra predictor 31b will be explained. Also at the intra predictor 31b, four prediction modes of Vertical mode, Horizontal mode, DC mode and Plane prediction mode exist in the intra color difference prediction mode. The intra predictor 31b serves to perform prediction with 8×16 block constituted by collecting successive two 8×8 blocks in a longitudinal direction within macro block being as unit with respect to an image signal in which chroma format signal indicates 4:2:2 format and color space signal indicates YCbCr. The techniques of generating prediction images in accordance with respective four prediction modes at the intra predictor 31b will be explained below.

(30) (a) Vertical mode (mode=0)

(31) In the Vertical mode, pixels of adjacent upper side block of color difference block are copied to allow the pixels thus copied to be prediction image of corresponding block. When pixels of adjacent upper side block, are expressed as p[x, −1], prediction image predc of color difference in this case is represented by the following formula (26), it is to be noted that this mode can be used only in the case where adjacent upper side block exists.
[15]
pred.sub.c[x,y]=p[x,−1](x=0 . . . 7,y=0 . . . 15)  (26)

(32) (b) Horizontal mode (mode=1)

(33) In the Horizontal mode, pixels of adjacent left side block of color difference block are copied to allow the pixels thus copied to be prediction image of corresponding block. When pixels of adjacent left side block are expressed as p[−1, y], prediction image predc of the color difference block in this case is represented by the following formula (27). It is to be noted that this mode can be used only in the case where adjacent left side block exists.
[16]
pred.sub.c[x,y]=p[−1,y](x=0 . . . 7,y=0 . . . 15)  (27)

(34) (c) DC mode (mode=2)

(35) In the DC mode, pixels of adjacent upper and left side blocks of color difference block are used to allow the mean (average) value thereof to be prediction image. It is to be noted that in the case where adjacent pixels do not exist, value 128 is used as prediction signal.

(36) Namely, in the case of x, y=0 . . . 3, prediction image predc [x, y] is generated by using upper side pixel p[x, −1] and left side pixel p[−1, y] which are adjacent (in this example, x, y=0 . . . 3). More particularly, in four cases of (i) the case where pixel p[x, −1] and p[−1, y] both exist, (ii) the case where pixel p[x, −1] exists and pixel p[−1, y] does not exist, (iii) the case where pixel p[x, −1] does not exist and pixel p[−1, y] exists, and (iv) the case where pixel p[x, −1] and pixel p[−1, y] do not both exist, prediction images are respectively generated in accordance with the following formulas (28) to (31).

(37) 0$\begin{array}{cc}\left[17\right]& \\ {\mathrm{pred}}_c\left[x,y\right]=\left(\underset{x^{\prime }=0}{\overset{3}{.Math.}}p\left[x^{\prime },-1\right]+\underset{y^{\prime }=0}{\overset{3}{.Math.}}p\left[-1,y^{\prime }\right]+4\right)>>3\left(x,y=0\mathrm{.Math.}3\right)& \left(28\right)\\ {\mathrm{pred}}_c\left[x,y\right]=\left(\underset{x^{\prime }=0}{\overset{3}{.Math.}}p\left[x^{\prime },-1\right]+2\right)>>2\left(x,y=0\mathrm{.Math.}3\right)& \left(29\right)\\ {\mathrm{pred}}_c\left[x,y\right]=\left(\underset{y^{\prime }=0}{\overset{3}{.Math.}}p\left[-1,y^{\prime }\right]+2\right)>>2\left(x,y=0\mathrm{.Math.}3\right)& \left(30\right)\end{array}$

(38) Similarly, in the case of x=4 . . . 7, y=0 . . . 3, prediction age predc [x, y] is generated by using upper side pixel p[x, −1] and left side pixel p[−1, y] which are adjacent (in this example, x=4 . . . 7, y=0 . . . 3). More particularly, in three cases of (i) the case where pixel p[x, −1] exists, (ii) the case where pixel p[x, −1] does not exist and pixel p[−1, y] exists, and (iii) the case where pixel p[x, −1] and pixel p[−1, y] do not both exist, prediction images are respectively generated in accordance with the following formulas (32) to (34).

(39) $\begin{array}{cc}\left[18\right]& \\ {\mathrm{pred}}_c\left[x,y\right]=\left(\underset{x^{\prime }=0}{\overset{3}{.Math.}}p\left[x^{\prime },-1\right]+2\right)>>2\left(x=4\mathrm{.Math.}7,y=0\mathrm{.Math.}3\right)& \left(32\right)\\ {\mathrm{pred}}_c\left[x,y\right]=\left(\underset{y^{\prime }=0}{\overset{3}{.Math.}}p\left[-1,y^{\prime }\right]+2\right)>>2\left(x=4\mathrm{.Math.}7,y=0\mathrm{.Math.}3\right)& \left(33\right)\\ {\mathrm{pred}}_c\left[x,y\right]=128\left(x=4\mathrm{.Math.}7,y=0\mathrm{.Math.}3\right)& \left(34\right)\end{array}$

(40) Similarly, in the case of x=0 . . . 3, y=4 . . . 7, prediction image predc [x, y] is generated by using upper side pixel p[x, −1] and left side pixel p[−1,y] which are adjacent (in this example, x=0 . . . 3, y=4 . . . 7). More particularly, in three cases of (i) the case where pixel p[−1, y] exists, (ii) the case where pixel p[x, −1] exists and pixel p[−1, y] does not exist, and (iii) the case where pixel p[x, −1] and p[−1, y] do not both exist, prediction images are respectively generated in accordance with the following formulas (35) to (37).

(41) $\begin{array}{cc}\left[19\right]& \\ {\mathrm{pred}}_c\left[x,y\right]=\left(\underset{y^{\prime }=4}{\overset{7}{.Math.}}p\left[-1,y^{\prime }\right]+2\right)>>2\left(x=0\mathrm{.Math.}3,y=4\mathrm{.Math.}7\right)& \left(35\right)\\ {\mathrm{pred}}_c\left[x,y\right]=\left(\underset{x^{\prime }=0}{\overset{3}{.Math.}}p\left[x^{\prime },-1\right]+2\right)>>2\left(x=0\mathrm{.Math.}3,y=4\mathrm{.Math.}7\right)& \left(36\right)\\ {\mathrm{pred}}_c\left[x,y\right]=128\left(x=0\mathrm{.Math.}3,y=4\mathrm{.Math.}7\right)& \left(37\right)\end{array}$

(42) Similarly, in the case of x, y=4 . . . 7, prediction image pre& [x, y] is generated by using upper side pixel p[x, −1] and left side pixel p[−1, y] which are adjacent (in this example, x, y=4 . . . 7). More particularly, in four cases of (i) the case where pixel p[x, −1] and pixel p[−1, y] both exist, (ii) the case where pixel p[x, −1] exists and pixel p[−1, y] does not exist, (iii) the case where pixel p[x, −1] does not exist and pixel p[−1, y] exists, and (iv) the case where pixel p[x, −1] and pixel p[−1, y] do not both exist, prediction images are respectively generated in accordance with the following formulas (38) to (41).

(43) $\begin{array}{cc}\left[20\right]& \\ {\mathrm{pred}}_c\left[x,y\right]=\left(\underset{x^{\prime }=4}{\overset{7}{.Math.}}p\left[x^{\prime },-1\right]+\underset{y^{\prime }=4}{\overset{7}{.Math.}}p\left[-1,y^{\prime }\right]+4\right)>>3\left(x,y=4\mathrm{.Math.}7\right)& \left(38\right)\\ {\mathrm{pred}}_c\left[x,y\right]=\left(\underset{x^{\prime }=4}{\overset{7}{.Math.}}p\left[x^{\prime },-1\right]+2\right)>>2\left(x,y=4\mathrm{.Math.}7\right)& \left(39\right)\\ {\mathrm{pred}}_c\left[x,y\right]=\left(\underset{y^{\prime }=4}{\overset{7}{.Math.}}p\left[-1,y^{\prime }\right]+2\right)>>2\left(x,y=4\mathrm{.Math.}7\right)& \left(40\right)\\ {\mathrm{pred}}_c\left[x,y\right]=128\left(x,y=4\mathrm{.Math.}7\right)& \left(41\right)\end{array}$

(44) Similarly, in the case of x=0 . . . 3, y=8 . . . 11, prediction image predc [x, y] is generated by using upper side pixel p[x, −1] and left side pixel p[−1, y] which are adjacent (in this example, x=0 . . . 3, y=8.011). More particularly, in three case of (i) the case where pixel p[−1, y] exists, (ii) the case where pixel p[x, −1] exists and pixel p[−1, y] does not exist, and (iii) the case where pixel p[x, −1] and pixel p[−1, y] do not both exist, prediction images are respectively generated in accordance with the following formulas (42) to (44).

(45) $\begin{array}{cc}\left[21\right]& \\ {\mathrm{pred}}_c\left[x,y\right]=\left(\underset{y^{\prime }=8}{\overset{11}{.Math.}}p\left[-1,y^{\prime }\right]+2\right)>>2\left(x=0\mathrm{.Math.}3,y=8\mathrm{.Math.}11\right)& \left(42\right)\\ {\mathrm{pred}}_c\left[x,y\right]=\left(\underset{x^{\prime }=0}{\overset{3}{.Math.}}p\left[x^{\prime },-1\right]+2\right)>>2\left(x=0\mathrm{.Math.}3,y=8\mathrm{.Math.}11\right)& \left(43\right)\\ {\mathrm{pred}}_c\left[x,y\right]=128\left(x=0\mathrm{.Math.}3,y=8\mathrm{.Math.}11\right)& \left(44\right)\end{array}$

(46) Similarly, in the case of x=4 . . . 7, y=8 . . . 11, prediction image predc [x, y] is generated by using upper side pixel p[x, −1] and left side pixel p[−1, y] which are adjacent (in this example, x=4 . . . 7, y=8 . . . 11). More particularly, in four cases of (i) the case where pixel p[x, −1] and pixel p[−1, y] both exist, (ii) the case where pixel p[x, −1] exists and pixel p[−1, y] does not exist, (iii) the case where pixel p[x, −1] does not exist and pixel p[−1, y] exists, and (iv) the case where pixel p[x, −1] and pixel p[−1, y] do not both exist, prediction images are respectively generated in accordance with the following formulas (45) to (48).

(47) $\begin{array}{cc}\left[22\right]& \\ {\mathrm{pred}}_c\left[x,y\right]=\left(\underset{x^{\prime }=4}{\overset{7}{.Math.}}p\left[x^{\prime },-1\right]+\underset{y^{\prime }=8}{\overset{11}{.Math.}}p\left[-1,y^{\prime }\right]+4\right)>>3\left(x=4\mathrm{.Math.}7,y=8\mathrm{.Math.}11\right)& \left(45\right)\\ {\mathrm{pred}}_c\left[x,y\right]=\left(\underset{x^{\prime }=4}{\overset{7}{.Math.}}p\left[x^{\prime },-1\right]+2\right)>>2\left(x=4\mathrm{.Math.}7,y=8\mathrm{.Math.}11\right)& \left(46\right)\\ {\mathrm{pred}}_c\left[x,y\right]=\left(\underset{y^{\prime }=8}{\overset{11}{.Math.}}p\left[-1,y^{\prime }\right]+2\right)>>2\left(x=4\mathrm{.Math.}7,y=8\mathrm{.Math.}11\right)& \left(47\right)\\ {\mathrm{pred}}_c\left[x,y\right]=128\left(x=4\mathrm{.Math.}7,y=8\mathrm{.Math.}11\right)& \left(48\right)\end{array}$

(48) Similarly, in the case of x=0 . . . 3, y=12 . . . 15, prediction image predc [x, y] is generated by using upper side pixel p[x, −1] and left side pixel p[−1, y] which are adjacent (in this example, x=0 . . . 3, y=12 . . . 15). More particularly, in three cases of (i) the case where pixel p[−1, y] exists, (ii) the case where pixel p[x, −1] exists and pixel p[−1, y] does not exist, and (iii) the case where pixel p[x, −1] and pixel p[−1, y] do not both exist, prediction images are respectively generated in accordance with the following formulas (49) to (51)

(49) $\begin{array}{cc}\left[23\right]& \\ {\mathrm{pred}}_c\left[x,y\right]=\left(\underset{y^{\prime }=12}{\overset{15}{.Math.}}p\left[-1,y^{\prime }\right]+2\right)>>2\left(x=0\mathrm{.Math.}3,y=12\mathrm{.Math.}15\right)& \left(49\right)\\ {\mathrm{pred}}_c\left[x,y\right]=\left(\underset{x^{\prime }=12}{\overset{3}{.Math.}}p\left[x^{\prime },-1\right]+2\right)>>2\left(x=0\mathrm{.Math.}3,y=12\mathrm{.Math.}15\right)& \left(50\right)\\ {\mathrm{pred}}_c\left[x,y\right]=128\left(x=0\mathrm{.Math.}3,y=12\mathrm{.Math.}15\right)& \left(51\right)\end{array}$

(50) Similarly, in the case of x=4 . . . 3, y=12 . . . 15, prediction image predc [x, y] is generated by using upper side pixel p[x, −1] and left side pixel p[−1, y] which are adjacent (in this example, x=4 . . . 7, y=12 . . . 15). More particularly, in four cases of (i) the case where pixel p[x, −1] and pixel [−1, y] both exist, (ii) the case where pixel p[x, −1] exists and pixel p[−1, y] does not exist, (iii) the case where pixel p[x, −1] does not exist and pixel p[−1, y] exists, and (iv) the case where pixel p[x, −1] and pixel p[−1, y] do not both exist, prediction images are respectively generated in accordance with the following formulas (52) to (55).

(51) $\begin{array}{cc}\left[24\right]& \\ {\mathrm{pred}}_c\left[x,y\right]=\left(\underset{x^{\prime }=4}{\overset{7}{.Math.}}p\left[x^{\prime },-1\right]+\underset{y^{\prime }=12}{\overset{15}{.Math.}}p\left[-1,y^{\prime }\right]+4\right)>>3\left(x=4\mathrm{.Math.}7,y=12\mathrm{.Math.}15\right)& \left(52\right)\\ {\mathrm{pred}}_c\left[x,y\right]=\left(\underset{x^{\prime }=4}{\overset{7}{.Math.}}p\left[x^{\prime },-1\right]+2\right)>>2\left(x=4\mathrm{.Math.}7,y=12\mathrm{.Math.}15\right)& \left(53\right)\\ {\mathrm{pred}}_c\left[x,y\right]=\left(\underset{y^{\prime }=12}{\overset{15}{.Math.}}p\left[-1,y^{\prime }\right]+2\right)>>2\left(x=4\mathrm{.Math.}7,y=12\mathrm{.Math.}15\right)& \left(54\right)\\ {\mathrm{pred}}_c\left[x,y\right]=128\left(x=4\mathrm{.Math.}7,y=12\mathrm{.Math.}15\right)& \left(55\right)\end{array}$

(52) Here, in the above-described prediction method, since mean (average) value of eight pixels of upper side block and 16 pixels of left side block is simply caused to be prediction image, it is necessary to perform division by 24. Thus, there is the problem that operation quantity becomes many. In view of the above, the prediction method is modified in a manner as described below to perform division by 16 (=24), thereby making it possible to reduce operation quantity.

(53) Namely, in the case of x, y=0 . . . 7, prediction image prede [x, y] is generated by using upper side pixel p[x, −1] and left side pixel p[−1, y] which are adjacent (in this example, x, y=0 . . . 7). More particularly, in four cases of (i) the case where pixel p[x, −1] and pixel p[−1, y] both exist, (ii) the case where pixel p[x, −1] exists and pixel p[−1, y] does not exist, (iii) the case where pixel p[x, −1] does not exist and pixel p[−1, y] exists, and (iv) the case where pixel p[x, −1] and pixel p[−1, y] do not both exist, prediction images are respectively generated in accordance with the following formulas (56) to (59).

(54) $\begin{array}{cc}\left[25\right]& \\ {\mathrm{pred}}_c\left[x,y\right]=\left(\underset{x^{\prime }=0}{\overset{7}{.Math.}}p\left[x^{\prime },-1\right]+\underset{y^{\prime }=0}{\overset{7}{.Math.}}p\left[-1,y^{\prime }\right]+8\right)>>4\left(x,y=0\mathrm{.Math.}7\right)& \left(56\right)\\ {\mathrm{pred}}_c\left[x,y\right]=\left(\underset{x^{\prime }=0}{\overset{7}{.Math.}}p\left[x^{\prime },-1\right]+4\right)>>3\left(x,y=0\mathrm{.Math.}7\right)& \left(57\right)\\ {\mathrm{pred}}_c\left[x,y\right]=\left(\underset{y^{\prime }=0}{\overset{7}{.Math.}}p\left[-1,y^{\prime }\right]+4\right)>>3\left(x,y=0\mathrm{.Math.}7\right)& \left(58\right)\\ {\mathrm{pred}}_c\left[x,y\right]=128\left(x,y=0\mathrm{.Math.}7\right)& \left(59\right)\end{array}$

(55) Similarly, in the case of x=0 . . . 7, y=8 . . . 15, prediction image predc [x, y] is generated by using upper side pixel p[x, −1] and left side pixel p[−1, y] which are adjacent (in this example, x=0 . . . 7, y=8 . . . 15). More particularly, in three cases of (i) the case where pixel p[−1, y] exists, (ii) the case where pixel p[x, −1] exists and pixel p[−1, y] does not exist, and (iii) the case where pixel p[x, −1] and pixel p[−1, y] do not both exist, prediction images are respectively generated in accordance with the following formulas (60) to (62).

(56) $\begin{array}{cc}\left[26\right]& \\ {\mathrm{pred}}_c\left[x,y\right]=\left(\underset{y^{\prime }=8}{\overset{15}{.Math.}}p\left[-1,y^{\prime }\right]+4\right)>>3\left(x=0\mathrm{.Math.}7,y=8\mathrm{.Math.}15\right)& \left(60\right)\\ {\mathrm{pred}}_c\left[x,y\right]=\left(\underset{x^{\prime }=0}{\overset{7}{.Math.}}p\left[x^{\prime },-1\right]+4\right)>>3\left(x=0\mathrm{.Math.}7,y=8\mathrm{.Math.}15\right)& \left(61\right)\\ {\mathrm{pred}}_c\left[x,y\right]=128\left(x=0\mathrm{.Math.}7,y=8\mathrm{.Math.}15\right)& \left(62\right)\end{array}$

(57) (d) Plane Prediction mode (mode=3)

(58) In the Plane Prediction mode, prediction image is plane-approximated from pixel of left side block and pixel of upper side block which are adjacent of color difference block to allow the prediction image thus obtained to be prediction image of the corresponding block. When pixels of left and upper side blocks which are adjacent are respectively expressed as p[−1, y] and p[x, −1], prediction image precis of color difference in this case is represented by the following formula (63). Here, Clip1 in the formula (63) indicates that clipping is performed into the range from 0 to 255.

(59) 0$\begin{array}{cc}\left[27\right]& \\ \begin{array}{c}{\mathrm{pred}}_c\left[x,y\right]=\mathrm{Clip}1\left(\left(a+b\times \left(x-3\right)+c\times \left(y-7\right)+16\right)>>5\right)\\ \left(x=0\mathrm{.Math.}7,y=0\mathrm{.Math.}15\right)\\ \mathrm{where}\\ \{\begin{array}{c}a=16\times \left(p\left[-1,15\right]+p\left[7,-1\right]\right)\\ b=\left(17\times H+16\right)>>5\\ c=\left(5\times V+32\right)>>6\\ H=\underset{x^{\prime }=0}{\overset{3}{.Math.}}\left(x^{\prime }+1\right)\times \left(p\left[4+x^{\prime },-1\right]-p\left[2-x^{\prime },-1\right]\right)\\ V=\underset{y^{\prime }=0}{\overset{7}{.Math.}}\left(y^{\prime }+1\right)\times \left(p\left[-1,8+y^{\prime }\right]-p\left[-1,6-y^{\prime }\right]\right)\end{array}\end{array}& \left(63\right)\end{array}$

(60) (2-1-3)

(61) Subsequently, the operation of the intra predictor 31c will be explained. Also at the intra predictor 31c, four prediction modes of Vertical mode, Horizontal mode, DC mode and Plane prediction mode exist in the intra color difference prediction mode. The intra predictor 31c performs prediction with 16×16 block constituted by collecting four 8×8 blocks in longitudinal and lateral directions successive within macro block being as unit with respect to image signal in which chroma for signal indicates 4:4:4 format and color space signal indicates YCbCr, RGB or XYZ. Techniques of generating prediction images in accordance with respective four prediction modes at the intra predictor 31c will be explained.

(62) (a) Vertical mode (mode=0)

(63) In the Vertical mode, pixels of adjacent upper side block of color difference block are copied to allow the pixels thus copied to be prediction image of corresponding block. When pixels of adjacent upper side block are expressed as p[x, −1], prediction image predc of color difference in this case is represented by the following formula (64). It is to be noted that this mode can be used only in the case where adjacent upper side block exists.
[28]
pred.sub.c[x,y]=p[x,−1](x,y=0 . . . 15)  (64)

(64) (b) Horizontal mode (mode=1)

(65) In the Horizontal mode, pixels of adjacent left side block of color difference block are copied to allow the pixels thus copied to be prediction image of the corresponding block. When pixels of adjacent left side block are expressed as p[−1, y], prediction image predc of color difference block in this case is represented by the following formula (65). It is to be noted that this mode can be used only in the case where adjacent left side block exists.
[29]
pred.sub.c[x,y]=p[−1,y](x,y=(0 . . . 15)  (65)

(66) (c) DC mode (mode=2)

(67) In the DC mode, pixels of upper and lower side blocks which are adjacent of color difference block are used to allow the mean (average) value thereof to be prediction image. It is to be noted that in the case where adjacent pixels do not exist, value 128 is used as prediction signal.

(68) Namely, in the case of x, y=0 . . . 15, prediction image predc p[x, y] is generated by using upper side pixel p[x, −1] and left side pixel p[−1, y] which are adjacent (in this example, x, y=0 . . . 5). More particularly, in four cases of (i) the case where pixel p[x, −1] and pixel p[−1, y] both exist, (ii) the case where pixel p[x, −1] exists and pixel p[−1, y] does not exist, (iii) the case where pixel p[x, −1] does not exist and pixel p[−1, y] exists, and (iv) the case where pixel p[x, −1] and pixel p[−1, y] do not both exist, prediction images are respectively generated in accordance with the following formulas (66) to (69).

(69) $\begin{array}{cc}\left[30\right]& \\ {\mathrm{pred}}_c\left[x,y\right]=\left(\underset{x^{\prime }=0}{\overset{15}{.Math.}}p\left[x^{\prime },-1\right]+\underset{y^{\prime }=0}{\overset{15}{.Math.}}p\left[-1,y^{\prime }\right]+16\right)>>5\left(x,y=0\mathrm{.Math.}15\right)& \left(66\right)\\ {\mathrm{pred}}_c\left[x,y\right]=\left(\underset{x^{\prime }=0}{\overset{15}{.Math.}}p\left[x^{\prime },-1\right]+8\right)>>4\left(x,y=0\mathrm{.Math.}15\right)& \left(67\right)\\ {\mathrm{pred}}_c\left[x,y\right]=\left(\underset{y^{\prime }=0}{\overset{15}{.Math.}}p\left[-1,y^{\prime }\right]+8\right)>>4\left(x,y=0\mathrm{.Math.}15\right)& \left(68\right)\\ {\mathrm{pred}}_c\left[x,y\right]=128\left(x,y=0\mathrm{.Math.}15\right)& \left(69\right)\end{array}$

(70) (d) Plane Prediction mode (mode=3)

(71) In the Plane Prediction mode, prediction image is plane-approximated from pixel of left side block and pixel of upper side block which are adjacent of color difference block to allow the prediction image thus obtained to be prediction image of corresponding block. When pixels of left and upper side blocks which are adjacent are respectively expressed as p[−1, y] and p[x, −1], the prediction image predc of color difference in this case is represented by the following formula (70). Here, Clip1 in the formula (70) indicates that clipping into the range from 0 to 255 is performed.

(72) $\begin{array}{cc}\left[31\right]& \\ \begin{array}{c}{\mathrm{pred}}_c\left[x,y\right]=\mathrm{Clip}1\left(\left(a+b\times \left(x-7\right)+c\times \left(y-7\right)+16\right)>>5\right)\\ \left(x,y=0\mathrm{.Math.}15\right)\\ \mathrm{where}\\ \{\begin{array}{c}a=16\times \left(p\left[-1,15\right]+p\left[15,-1\right]\right)\\ b=\left(5\times H+32\right)>>6\\ c=\left(5\times V+32\right)>>6\\ H=\underset{x^{\prime }=0}{\overset{7}{.Math.}}\left(x^{\prime }+1\right)\times \left(p\left[8+x^{\prime },-1\right]-p\left[6-x^{\prime },-1\right]\right)\\ V=\underset{y^{\prime }=0}{\overset{7}{.Math.}}\left(y^{\prime }+1\right)\times \left(p\left[-1,8+y^{\prime }\right]-p\left[-1,6-y^{\prime }\right]\right)\end{array}\end{array}& \left(70\right)\end{array}$

(73) (2-2) Orthogonal Transform Unit

(74) Chroma format signal and color space signal are also delivered to the orthogonal transform unit 14.

(75) One example of the configuration of the orthogonal transform unit 14 is shown in FIG. 7. The orthogonal transform unit 14 switches orthogonal transform system on the basis of chroma format signal indicating whether resolution of color component is that of any one of the 4:2:0 format, the 4:2:2 format and the 4:4:4 format, etc., and color space signal indicating whether color space is any one of YCbCr, RGB and XYZ, etc.

(76) At the orthogonal transform unit 14 shown in FIG. 7, the chroma format signal and the color space signal are delivered to switches 40, 42. The switches 40, 42 select any one of orthogonal transform elements 41a, 41b, 41c on the basis of the chroma format signal and the color space signal to deliver output from the adder 13 to the selected orthogonal transform element to output a signal from the selected orthogonal transform element. The switches 40, 42 select the same orthogonal transform element. It is to be noted that while explanation will be given in FIG. 7 on the premise that any one of three kinds of orthogonal transform elements 41a, 41b, 41c is selected, the number of orthogonal transform elements, i.e., the number of orthogonal transform systems may be arbitrarily set.

(77) (2-2-1)

(78) First, the operation of the orthogonal transform element 41a will be explained. The orthogonal transform element 41a performs orthogonal transform with respect to an image signal in which chroma format signal indicates 4:2:0 format and color space signal indicates YCbCr. It is to be noted that since the operation of the orthogonal transform element 41a is the same as that of the previously described prior art, the detailed explanation thereof is omitted.

(79) (2-2-2)

(80) Then, the operation of the orthogonal transform element 41b will be explained. The orthogonal transform element 41b performs orthogonal transform with respect to an image signal in which chroma format signal indicates 4:2:2 format and color space signal indicates YCbCr.

(81) More particularly, after intra-prediction of color difference is performed, 4×4 integer transform is applied on 4×4 pixel block basis within 8×8 blocks. When difference signal obtained by subtracting prediction image from corresponding pixel block is expressed as f4×4, 4×4 orthogonal transform processing is represented by the following formula (71).

(82) $\begin{array}{cc}\left[32\right]& \\ \begin{array}{c}\begin{array}{c}{f}_{4\times 4}=T_{4\times 4}\times F_{4\times 4}\times T_{4\times 4}^T\\ \mathrm{where}\end{array}\\ T_{4\times 4}=\left(\begin{array}{cccc}1& 1& 1& 1\\ 2& 1& -1& -2\\ 1& -1& -1& 1\\ 1& -2& 2& -1\end{array}\right)\end{array}& \left(71\right)\end{array}$

(83) After 4×4 integer transform processing is performed, (0, 0) coefficients of eight 4×4 blocks within two 8×8 blocks successive in a longitudinal direction are collected to constitute 2×4 block to apply 2×4 transform processing to the 2×4 block. This is because efficiency of intra-prediction used in color difference is not so high so that correlation is still left between (0, 0) coefficients of adjacent 4×4 blocks. In order to further enhance (increase) encoding efficiency by making use of the correlation, only (0, 0) coefficients of 4×4 blocks are collected to constitute 2×4 blocks to apply 2×4 transform processing thereto. When block of chroma DC of 2×4 is expressed as fdc 2×4, transform processing with respect to the chroma DC block is represented by the following formula (72),

(84) $\begin{array}{cc}\left[33\right]& \\ \begin{array}{c}\begin{array}{c}{\mathrm{fdc}}_{2\times 4}^{\prime }=T_{v\left(4\right)}\times {\mathrm{fdc}}_{2\times 4}\times T_{h\left(2\right)}^T\\ \mathrm{where}\end{array}\\ T_{v\left(4\right)}=\left(\begin{array}{cccc}1& 1& 1& 1\\ 1& 1& -1& -1\\ 1& -1& -1& 1\\ 1& -1& 1& -1\end{array}\right)\\ T_{h\left(2\right)}=\left(\begin{array}{cc}1& 1\\ 1& -1\end{array}\right)\end{array}& \left(72\right)\end{array}$

(85) (2-2-3)

(86) Subsequently, the operation of the orthogonal transform element 41c will be explained. The orthogonal transform element 41c performs orthogonal transform with respect to an image signal in which chroma format signal indicates 4:4:4 format and color space signal indicates YCbCr, RGB or XYZ.

(87) More particularly, 4×4 integer transform of color difference indicating 4:4:4 format, YCbCr, RGB or XYZ is performed thereafter to collect 16 (0, 0) coefficients within macro block in the same manner as the case of luminance to constitute 4×4 DC block to apply 4×4 transform processing thereto. This transform processing is represented by the following formula (73).

(88) $\begin{array}{cc}\left[34\right]& \\ \begin{array}{c}\begin{array}{c}{\mathrm{fdc}}_{4\times 4}^{\prime }=T_{4\times 4}\times {\mathrm{fdc}}_{4\times 4}\times T_{4\times 4}^T\\ \mathrm{where}\end{array}\\ T_{4\times 4}=\left(\begin{array}{cccc}1& 1& 1& 1\\ 1& 1& -1& -1\\ 1& -1& -1& 1\\ 1& -1& 1& -1\end{array}\right)\end{array}& \left(73\right)\end{array}$

(89) (2-3) Quantization Unit

(90) Chroma format signal and color space signal are also delivered to the quantization unit 15.

(91) An example of the configuration of the quantization unit 15 is shown in FIG. 9. The quantization unit 15 switches quantization system on the basis of chroma format signal indicating whether resolution of color component is that of any one of 4:2:0 format, 4:2:2 format and 4:4:4 format, etc. and color space signal indicating whether color space is any one of YCbCr, RGB and XYZ, etc.

(92) At the quantization unit 15 shown in FIG. 9, chroma format signal and color space signal are delivered to switches 50, 52. The switches 50, 52 select any one of quantizers 51a, 51b, 51c on the basis of chroma format signal and color space signal to deliver an output from the orthogonal transform unit 14 to the selected quantizer to output a signal from the selected quantizer. The switches 50, 52 select the same quantizer. It is to be noted that while explanation will be given in FIG. 9 on the premise that any one of three kinds of quantizers 51a, 51b, 51c is selected, the number of quantizers, i.e., the number of quantization systems may be arbitrarily set.

(93) (2-3-1)

(94) First, the operation of the quantizer 51a will be explained. The quantizer 51a performs quantization with respect to an image signal in which chroma format signal indicates 4:2:0 format and color space signal indicates YCbCr. It is to be noted that since the operation of the quantizer 51a is the same as that of the previously described prior art, the detailed explanation thereof is omitted.

(95) (2-3-2)

(96) Then, the operation of the quantizer 51b will be explained. The quantizer 51b performs quantization with respect to an image signal in which chroma format signal indicates 4:2:2 format and color space signal indicates YCbCr.

(97) Here, Hadamard transform used in transform processing of chroma DC in the case of 4:2:0 format is represented by the following formula (74).

(98) $\begin{array}{cc}\left[35\right]& \\ \begin{array}{c}\begin{array}{c}\begin{array}{c}{\mathrm{fdc}}_{2\times 4}^{\prime }=T_2\times {\mathrm{fdc}}_{2\times 2}\times T_2^T\\ =\frac{1}{2}\left(\begin{array}{cc}1& 1\\ 1& -1\end{array}\right){\mathrm{fdc}}_{2\times 2}\left(\begin{array}{cc}1& 1\\ 1& -1\end{array}\right)\end{array}\\ \mathrm{where}\end{array}\\ T_2=\frac{1}{\sqrt{2}}\left(\begin{array}{cc}1& 1\\ 1& -1\end{array}\right)\end{array}& \left(74\right)\end{array}$

(99) On the other hand, 2×4 transform used in transform processing of chroma DC in the case of 4:2:2 format is represented by the following formula (75).

(100) $\begin{array}{cc}\left[36\right]& \\ \begin{array}{c}\begin{array}{c}\begin{array}{c}{\mathrm{fdc}}_{2\times 4}^{\prime }=T_{v\left(4\right)}\times {\mathrm{fdc}}_{2\times 4}\times T_{h\left(2\right)}^T\\ =\frac{1}{2\sqrt{2}}\left(\begin{array}{cccc}1& 1& 1& 1\\ 1& 1& -1& -1\\ 1& -1& -1& 1\\ 1& -1& 1& -1\end{array}\right)f_{2\times 4}\left(\begin{array}{cc}1& 1\\ 1& -1\end{array}\right)\end{array}\\ \mathrm{where}\end{array}\\ T_{v\left(4\right)}=\left(\begin{array}{cccc}1& 1& 1& 1\\ 1& 1& -1& -1\\ 1& -1& -1& 1\\ 1& -1& 1& -1\end{array}\right)\\ T_{h\left(2\right)}=\left(\begin{array}{cc}1& 1\\ 1& -1\end{array}\right)\end{array}& \left(75\right)\end{array}$

(101) Accordingly, normalization coefficient by transform processing in the 4:2:0 format is ½, whereas normalization coefficient by transform processing in the 4:2:2 format is ½√2. However, since real number operation is included in this case, 2×4 transform is simplified as indicated by the following formula (76).

(102) $\begin{array}{cc}\left[37\right]& \\ \begin{array}{c}{\mathrm{fdc}}_{2\times 4}^{\prime }=T_{v\left(4\right)}\times {\mathrm{fdc}}_{2\times 4}\times T_{h\left(2\right)}^T\\ =\frac{1}{2\sqrt{2}}\left(\begin{array}{cccc}1& 1& 1& 1\\ 1& 1& -1& -1\\ 1& -1& -1& 1\\ 1& -1& 1& -1\end{array}\right){\mathrm{fdc}}_{2\times 4}\left(\begin{array}{cc}1& 1\\ 1& -1\end{array}\right)\\ \approx \frac{1}{4}\left(\begin{array}{cccc}1& 1& 1& 1\\ 1& 1& -1& -1\\ 1& -1& -1& 1\\ 1& -1& 1& -1\end{array}\right){\mathrm{fdc}}_{2\times 4}\left(\begin{array}{cc}1& 1\\ 1& -1\end{array}\right)\end{array}& \left(76\right)\end{array}$

(103) Since the normalization coefficients are calculated together with scale in quantization, it is necessary to change the quantization method in a manner as described below in the case of transform processing of 4:2:2 format.

(104) When quantized DC coefficient is Qf′[ij], quantized coefficient values of 2×4 chroma DC block are given by, e.g., the following formula (77). Here, r in the formula (77) is parameter for changing rounding processing. It is to be noted that since quantization with respect to AC coefficients is the same as that in the case of the 4:2:0 format, the explanation thereof will be omitted.
[38]
Qfdc′[i,j]=(fdc′[i,j]×Q(QP.sub.c%6,0,0)+r)>>(15+QPc/6)(i=0 . . . 1,j=0 . . . 3)  (77)

(105) (2-3-3)

(106) Subsequently, the operation of the quantizer 51c will be explained. The quantizer 51c performs quantization with respect to an image signal in which chroma format signal indicates 4:4:4 format and color space signal indicates YCbCr, RGB or XYZ.

(107) Here, Hadamard transform used in transform processing of chroma DC is represented by the following formula (78). Accordingly, in this case, the normalization coefficient of transform processing becomes equal to ¼.

(108) $\begin{array}{cc}\left[39\right]& \\ \begin{array}{c}\begin{array}{c}{\mathrm{fdc}}_{4\times 4}^{\prime }=T_4\times {\mathrm{fdc}}_{4\times 4}\times T_4^T\\ =\frac{1}{4}\left(\begin{array}{cccc}1& 1& 1& 1\\ 1& 1& -1& -1\\ 1& -1& -1& 1\\ 1& -1& 1& -1\end{array}\right){\mathrm{fdc}}_{4\times 4}\left(\begin{array}{cccc}1& 1& 1& 1\\ 1& 1& -1& -1\\ 1& -1& -1& 1\\ 1& -1& 1& -1\end{array}\right)\end{array}\\ \mathrm{where}\\ T_4=\frac{1}{2}\left(\begin{array}{cccc}1& 1& 1& 1\\ 1& 1& -1& -1\\ 1& -1& -1& 1\\ 1& -1& 1& -1\end{array}\right)\end{array}& \left(78\right)\end{array}$

(109) When quantized DC coefficient is Qf′[i j], quantized coefficient value of 4×4 chroma DC block is given by, e.g., the following formula (79). Here, r in the formula (79) is parameter for changing rounding processing.
[40]
Qfdc′[i,j]=(fdc′[i,j]×Q(QP.sub.c%6,0,0)+r)>>(15+QP.sub.c/6)(i,j=0 . . . 3)  (79)

(110) (2-4) Inverse Quantization Unit

(111) Chroma format signal and color space signal are also delivered to the inverse quantization unit 18.

(112) One example of the configuration of the inverse quantization unit 18 is shown in FIG. 10. The inverse quantization unit 18 switches inverse quantization system on the basis of chroma format signal indicating whether resolution of color component is that of any one of 4:2:0 format, 4:2:2 format and 4:4:4 format, etc. and color space signal indicating whether color space is any one of YCbCr, RGB and XYZ, etc.

(113) In the inverse quantization unit 18 shown in FIG. 10, chroma format signal and color space signal are delivered to switches 60, 62. The switches 60, 62 select any one of inverse quantizers 61a, 61b, 61c on the basis of the chroma format signal and the color space signal to deliver output from the quantization unit 15 to the selected inverse-quantizer to output a signal from the selected inverse-quantizer. The switches 60, 62 select the same inverse-quantizer. It is to be noted that while explanation will be given in the FIG. 10 on the premise that any one of three kinds of inverse-quantizers 61a, 61b, 61c is selected, the number of inverse-quantizers, i.e., the number of inverse-quantization systems may be arbitrarily set.

(114) (2-4-1)

(115) First, the operation of the inverse-quantizer 61a will be explained. The inverse-quantizer 6a performs inverse-quantization with respect to an image signal in which chroma format signal indicates 4:2:0 format and color space signal indicates YCbCr. It is to be noted that since the operation of the inverse-quantizer 61a is the same as that of the previously described prior art, the detailed explanation thereof will be omitted.

(116) (2-4-2)

(117) Then, the operation of the inverse-quantizer 61b will be explained. The inverse-quantizer 61b performs inverse quantization with respect to an image signal in which chroma format signal indicates 4:2:2 format and color space signal indicates YCbCr.

(118) More particularly, when inverse-quantized DC coefficient is fdc″, inverse-quantized DC coefficient value of 2×2 chroma DC block is represented by the following formula (80) in the case where QPc is 6 (six) or more, and is represented by the following formula (81) in the case where QPc is less than 6 (six). It is to be noted that since inverse-quantization with respect to AC coefficients is the same as that in the case of 4:2:0 format, the explanation thereof will be omitted.
[41]
fdc″[i,j]=(fdc′[i,j]×Q(QP.sub.c%6,0,0))<<(QP.sub.c/6−2)(i=0 . . . 1,j=0 . . . 3)  (80)
fdc″[i,j]=(fdc′[i,j]×Q(QP.sub.c%6,0,0))>>(2−QP.sub.c/6)(i=0, . . . 1,j=0 . . . 3)  (81)

(119) (2-4-3)

(120) Then, the operation of the inverse-quantizer 61c will be explained. The inverse-quantizer 61c performs inverse quantization with respect to an image signal in which chroma format signal indicates 4:4:4 format and color space signal indicates YCbCr, RGB or XYZ.

(121) More particularly, when inverse-quantized DC coefficient is fdc″, inverse-quantized coefficient value of 4×4 chroma DC block is represented by the following formula (82) in the case where QPc is 6 (six) or more, and is represented by the following formula (83) in the case where QPc is less than 6 (six). It is to be noted that since inverse quantization with respect to AC coefficients is the same as that in the case of 4:2:0 format, the explanation thereof will be omitted.
[42]
fdc″[i,j]=(fdc′[i,j]×Q(QP.sub.c%6,0,0))<<(QP.sub.c/6−2)(i,j=0 . . . 3)  (82)
fdc″[i,j]=(fdc′[i,j]×Q(QP.sub.c%6,0,0))>>(2−QP.sub.c/6)(i,j=0 . . . 3)  83)

(122) (2-5) Inverse Orthogonal Transform Unit

(123) Chroma format signal and color space signal are also delivered to the inverse orthogonal transform unit 19.

(124) One example of the configuration of the inverse orthogonal transform unit 19 is shown in FIG. 11. The inverse orthogonal transform unit 19 switches inverse orthogonal transform system on the basis of chroma format signal indicating whether resolution of color component is that of any one of 4:2:0 format, 4:2:2 format and 4:4:4 format, etc. and color space signal indicating whether color space is any one of YCbCr, RGB and XYZ, etc.

(125) In the inverse orthogonal transform unit 19 shown in FIG. 11, chroma format signal and color space signal are delivered to switches 70, 72. The switches 70, 72 select any one of inverse orthogonal transform elements 71a, 71b, 71c on the basis of the chroma format signal and the color space signal to deliver an output from the inverse quantization unit 18 to the selected inverse orthogonal transform element to output a signal from the selected inverse orthogonal transform element. The switches 70, 72 select the same inverse orthogonal transform element. It is to be noted that while explanation will be given in the FIG. 11 on the premise that any one of three kinds of inverse orthogonal transform elements 71a, 71b, 71c is selected, the number of inverse orthogonal transform elements, i.e., the number of inverse orthogonal transform systems may be arbitrarily set.

(126) (2-5-1)

(127) First, the operation of the inverse orthogonal transform element 71a will be explained. The inverse-orthogonal transform element 71a performs inverse orthogonal transform with respect to an image signal in which chroma format signal indicates 4:2:0 format and color space signal indicates YCbCr. It is to be noted that since the operation of the inverse orthogonal transform element 71a is the same as that of the previously described prior art, the detailed explanation thereof will be omitted.

(128) (2-5-2)

(129) Then, the operation of the inverse orthogonal transform element 71b will be explained. The inverse orthogonal transform element 71b performs inverse orthogonal transform with respect to an image signal in which chroma format signal indicates 4:2:2 format and color space signal indicates YCbCr.

(130) More particularly, 2×4 inverse transform processing is applied to 2×4DC block. When inverse-transformed 2×4 chroma DC block is expressed as fdc2×4′″, inverse transform with respect to the chroma DC block is represented by the following formula (84).

(131) 0$\begin{array}{cc}\left[43\right]& \\ \begin{array}{c}{\mathrm{fdc}}_{2\times 4}^{\mathrm{\prime \prime \prime }}=T_{v\left(4\right)}\times {\mathrm{fdc}}_{2\times 4}^{″}\times T_{h\left(2\right)}^T\\ \mathrm{where}\\ T_{v\left(4\right)}=\left(\begin{array}{cccc}1& 1& 1& 1\\ 1& 1& -1& -1\\ 1& -1& -1& 1\\ 1& -1& 1& -1\end{array}\right)\\ T_{h\left(2\right)}=\left(\begin{array}{cc}1& 1\\ 1& -1\end{array}\right)\end{array}& \left(84\right)\end{array}$

(132) With the chroma DC coefficient being as (0, 0) coefficients of 4×4 block as shown in FIG. 8, inverse transform processing of respective 4×4 blocks is performed. When respective coefficients of 4×4 blocks in which fdc2×4′″ which is inverse-transformed chroma DC is caused to be (0, 0) coefficient are expressed as F′4×4 and decoded difference signal at inverse transformed 4×4 block is expressed as F″4×4, inverse transform processing is represented by the following formula (85).

(133) $\begin{array}{cc}\left[44\right]& \\ \begin{array}{c}{F}_{4\times 4}^{″}=T_{4\times 4}\times F_{4\times 4}^{\prime }\times T_{4\times 4}^T\\ \mathrm{where}\\ T_{4\times 4}=\left(\begin{array}{cccc}1& 1& 1& 1\\ 2& 1& -1& -2\\ 1& -1& -1& 1\\ 1& -2& 2& -1\end{array}\right)\end{array}& \left(85\right)\end{array}$

(134) (2-5-3)

(135) Subsequently, the operation of the inverse orthogonal transform element 71c will be explained. The inverse orthogonal transform element 71c performs inverse orthogonal transform with respect to an image signal in which chroma format signal indicates 4:4:4 format and color space signal indicates YCbCr, RGB or XYZ.

(136) More particularly, 4×4 inverse transform processing is applied to 4×4 DC blocks. When inverse-transformed 4×4 chroma DC block is expressed as fdc4×4′″, inverse transform processing with respect to the chroma DC block is represented by the following formula (86).

(137) $\begin{array}{cc}\left[45\right]& \\ \begin{array}{c}{\mathrm{fdc}}_{4\times 4}^{\mathrm{\prime \prime \prime }}=T_4\times {\mathrm{fdc}}_{4\times 4}^{″}\times T_4^T\\ \mathrm{where}\\ T_4=\left(\begin{array}{cccc}1& 1& 1& 1\\ 1& 1& -1& -1\\ 1& -1& -1& 1\\ 1& -1& 1& -1\end{array}\right)\end{array}& \left(86\right)\end{array}$

(138) With this aroma DC coefficient being as (0, 0) coefficient of 4×4 block of AC coefficients, inverse transform processing of respective 4×4 blocks is performed. When respective coefficients of 4×4 blocks in which fdc4×4′″ which is inverse-transformed chroma DC is caused to be (0, 0) coefficient are expressed as F′4×4, and decoded difference signal at inverse-transformed 4×4 block is expressed as F″4×4, inverse transform processing is represented by the following formula (87).

(139) $\begin{array}{cc}\left[46\right]& \\ \begin{array}{c}{F}_{4\times 4}^{″}=T_{4\times 4}\times F_{4\times 4}^{\prime }\times T_{4\times 4}^T\\ \mathrm{where}\\ T_{4\times 4}=\left(\begin{array}{cccc}1& 1& 1& 1\\ 2& 1& -1& -2\\ 1& -1& -1& 1\\ 1& -2& 2& -1\end{array}\right)\end{array}& \left(87\right)\end{array}$

(140) (2-6) Other Block

(141) The chroma format signal and the color space signal are also delivered to the reversible encoding unit 16, at which variable length encoding or arithmetic encoding of such signals is performed. The signals thus obtained are outputted in the state included in image compressed information.

(142) The chroma format signal and the color space signal are encoded by, e.g., syntax as described below.

(143) $\mathrm{seq\_parameter}\mathrm{\_set}\mathrm{\_rbsp}\left(\right)\{\mathrm{.Math.}\mathrm{chroma\_format}\mathrm{\_idc}u\left(2\right)\mathrm{color\_space}\mathrm{\_idc}u\left(2\right)\mathrm{.Math.}\}$

(144) Here, syntax encoded as u(2) is encoded by variable length code of, e.g., “001x1x0”. Among them, x1 and x0 correspond to 2 (two) bits of syntax to be encoded,

(145) (3) Configuration and Operation of the Image Information Decoding Apparatus

(146) Outline of the configuration of an image information decoding apparatus corresponding to the above-described image information encoding apparatus 10 is shown in FIG. 12. As shown in FIG. 12, the image information decoding apparatus 80 comprises a storage buffer 81, a reversible decoding unit 82, an inverse quantization unit 83, an inverse orthogonal transform unit 84, an adder 85, an image sorting buffer 86, a D/A (Digital/Analogue) converting unit 87, a motion prediction/compensation unit 88, a frame memory 89, and an intra prediction unit 90.

(147) In FIG. 12, an image compressed information serving as input is first stored into the storage buffer 81, and is then transferred to the reversible decoding unit 82. The reversible decoding unit 82 performs processing such as variable length decoding or arithmetic decoding, etc. on the basis of a predetermined format for image compressed information. Moreover, in the case where corresponding frame is inter-encoded frame, the reversible decoding unit 82 also decodes motion vector information stored at header portion of the image compressed information to transfer the decoded information thus obtained to the prediction/compensation unit 88. Further, the reversible decoding unit 82 decodes chroma format signal and color space signal to deliver decoded signals thus obtained to the inverse quantization unit 83, the inverse orthogonal transform unit 84 and the intra prediction unit 90.

(148) Quantized transform coefficients serving as output of the reversible decoding unit 82 are delivered to the inverse quantization unit 83, at which they are outputted as transform coefficients. The inverse orthogonal transform unit 84 implements reversible transform such as inverse discrete cosine transform or inverse Karhunen-Loeve transform, etc. to the transform coefficients on the basis of a predetermined format for image compressed information. In the case where corresponding frame is intra-encoded frame, image information to which inverse orthogonal transform processing has been implemented is stored into the image sorting buffer 86, and is outputted after undergone D/A converting processing.

(149) Here, in the case where corresponding frame or macro block is intra-encoded frame or macro block, decoding processing is performed by using the same inverse quantization method, inverse orthogonal transform method and intra prediction method as those as described above on the basis of the chroma format signal and the color space signal which have been decoded at the reversible decoding unit 82.

(150) On the other hand, in the case where corresponding frame is inter-encoded frame, reference image is generated on the basis of motion vector information to which reversible decoding processing has been implemented and image information stored in the frame memory 89. The reference image thus generated and output of the inverse orthogonal transform unit 84 are synthesized at the adder 85. Since other processing are the same as those of intra-encoded frame, the explanation thereof will be omitted.

(151) It is to be noted that while the present invention has been described in accordance with certain preferred embodiments thereof illustrated in the accompanying drawings and described in the above description in detail, it should be understood by those ordinarily skilled in the art that the invention is not limited to embodiments, but various modifications, alternative constructions or equivalents can be implemented without departing from the scope and spirit of the present invention as set forth by appended claims.

INDUSTRIAL APPLICABILITY

(152) The present invention can efficiently perform encoding processing by using intra-image predictive encoding processing not only with respect to the case of input image signal in which corresponding frame is 4:2:0 format and color space is YCbCr, but also with respect to the case of input image signal in which corresponding format is 4:2:2 format or 4:4:4 format, and color space is RGB or XYZ, etc.