JPH10191393A - Multi-view-point image coder - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve the compression rate while keeping a high compression rate by applying motion compression only without parallax compression when no motion amount is available in a 2nd motion compression means so as to reduce the entire information amount of a vector required for coding. SOLUTION: A parallax compression section 2 applies compression by block matching with respect to a right eye image R1 by using a decoded image of a left eye image L1. The parallax vector in this case takes difference of a block extracted from a block composite image of right and left eye images R1, L1 and a vector where a total sum (SAD) of errors in a difference block is selected. With respect to a block of a 2nd right eye image R2, a motion compression section 14 uses the preceding right eye image R1 to conduct block matching and when a motion vector is (0, 0), only motion compression is conducted. In other cases, when the SAD is smaller than the total sum of the picture element in the block, the parallax compression is conducted and when not, parallax compression is not conducted.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to high-efficiency image coding, and more particularly to high-efficiency coding of multi-view images using motion compensation and parallax compensation prediction.

[0002]

2. Description of the Related Art In multi-viewpoint image coding, coding is performed by combining motion compensation generally used in moving image coding and disparity compensation between images having different viewpoints. Methods for improving coding efficiency are well known.

[0003] For example, in the literature (WA Schup, Yasuda, "
Data Compression of Stereo Video Using Parallax Compensation and Motion Compensation ", PCSJ88 (1988), pp. 63-6.
According to 4), as shown in FIG. 6, only the motion compensation is performed on the right image. With respect to the left image, a method of selecting the compensation method having the smaller prediction error from the parallax compensation and the motion compensation for each block and coding the selected image.

[0004]

However, in the conventional method, only the parallax compensation or only the motion compensation is performed, so that the temporal redundancy remains when only the parallax compensation is performed, and the spatial redundancy remains when the motion compensation is performed alone. Inefficient compression.

[0005] In addition, since switching is performed so that only parallax compensation or only motion compensation is performed for each block, it is different for each block whether a vector obtained by compensation is a parallax vector or a motion vector having different properties. There is a problem that the encoding efficiency of the vector is deteriorated.

An object of the present invention is to realize high-efficiency encoding by combining motion compensation and disparity compensation prediction in encoding a multi-view image, and to achieve matching accuracy by using variance as a criterion for the combined compensation. If there is no motion, the amount of vector information is reduced as a whole by performing only motion compensation, and the block is selected by referring to the compensation method of the surrounding blocks. An object of the present invention is to provide a multi-viewpoint encoding device capable of reducing distortion.

[0007]

In order to achieve the above object, a coding apparatus according to the present invention, in a multi-view image coding apparatus, determines an image of one viewpoint as a reference image, First motion compensation means for performing compensation, disparity compensation means for performing disparity compensation on a disparity image that is an image of a viewpoint other than the reference image, and disparity compensation for the disparity image or the disparity-compensated disparity image. Motion compensating means for performing motion compensation by means of a block, and a judging unit for selecting a compensation method for each block. When the second motion compensating means judges that there is no motion amount, no parallax compensation is performed. It is characterized in that only motion compensation is performed.

In the present invention, in the above-described parallax image encoding, when motion compensation is performed after performing parallax compensation or motion compensation or parallax compensation on a block having a block on the left, A predetermined value may be subtracted from the sum of the errors in the difference block by the same compensation as that of the block.

According to the present invention, in the encoding of the parallax image, after compensating all the blocks,
In a block in which a block near the surrounding 8 exists, the same compensation is performed for all the blocks near the surrounding 8 of the block, and the compensation performed in the block is performed for the block near the surrounding 8. When the compensation is different from the compensation, the compensation performed in the block may be matched with the compensation of a block near the surrounding 8.

Further, in the present invention, in the above-described parallax image encoding, when performing motion compensation after performing parallax compensation, a prediction error and a variance in a block are obtained, and a reference block having the smallest variance is used for prediction. May be used.

In the present invention, when encoding the parallax image, among the reference blocks having a variance equal to or less than a certain threshold, a reference block having the smallest sum of prediction errors in the block may be used for prediction.

Further, according to the present invention, when the total value and the variance of the errors in the block are ranked from the smallest one, weights are given to both the ranks, and when the sum of them is minimized. May be used for prediction.

In the present invention, when coding the parallax image, instead of performing motion compensation after performing parallax compensation, parallax compensation may be performed after performing motion compensation.

In the present invention, the prediction error in performing the motion compensation after performing the parallax compensation may be viewed in pixel units, and if any error is larger than a predetermined threshold, the motion compensation may not be performed.

In the present invention, when obtaining a disparity vector for each block, the disparity vector of the block determined not to perform the disparity compensation is set as the first disparity vector, and it is determined that the disparity compensation is performed before the first block. When there is a block, the second disparity vector of the block determined to perform the disparity compensation before the block is used instead of the first disparity vector, and the prediction error when the second disparity vector is used is reduced. If it is larger than a certain threshold value, it may have a first disparity vector.

[0016]

Embodiments of the present invention will be described below in detail. FIG. 1 shows an embodiment of a multi-view image encoding apparatus according to the present invention. The present embodiment is an example of a binocular system, and an image input to the image encoding device is a stereoscopic image having parallax. The three-dimensional images are assumed to be moving images.

First, the procedure for encoding the stereoscopic image 30 is shown in FIG.
This will be described with reference to FIG. Encoding is performed according to the following procedures (a) to (d).

(A) Encoding of left-eye image 31: The left-eye image 31 of the stereoscopic image 30 is encoded in a frame.

(B) Encoding of right-eye image 32: Inter-frame encoding is performed on the right-eye image 32 by parallax compensation using the decoded image of the left-eye image 31.

At this time, the disparity vectors are obtained for all the blocks, and it is determined whether or not the disparity compensation is performed for each block based on the sum of the errors in the blocks as in the normal motion compensation.

At the same time, the right eye image 3 is obtained from the disparity vector.
From 2, a difference image 33 is created by taking the difference between the decoded image of the left-eye image 31.

(C) Encoding of left-eye image 34... The left-eye image 34 is subjected to inter-frame encoding by motion compensation using a decoded image of the previous left-eye image 31.

(D) Encoding of right-eye image 35: Encoding of the right-eye image 35 is performed for each block, and will be described with reference to FIG.

The encoding is performed according to the following procedures (1) to (5).

(1) The right eye image 3 is determined by the determination unit 100 in FIG.
For block 5, after performing block matching using the decoded image of the right-eye image 32, if the prediction error is equal to or less than a certain threshold, motion compensation is performed; otherwise, it is determined not to perform motion compensation.

When the judgment section 100 judges that motion compensation is to be performed, 1 is substituted for the flag MC. When the judgment section 101 judges that the motion vector is (0, 0), the right-eye image 35 is added. On the other hand, a disparity vector is obtained using the decoded image of the left-eye image 34. The difference between the right-eye image 35 and the decoded image of the right-eye image 32 whose motion vector is (0, 0) motion-compensated is calculated, and the difference value is encoded.

(2) If the determination section 100 determines that motion compensation is not to be performed, 0 is substituted for the flag MC.

In both the case and the case where the decision unit 101 decides that the motion vector is not (0, 0), a disparity vector is calculated for the right-eye image 35 using the decoded image of the left-eye image 34. After that, the determination unit 102 determines that the parallax compensation is performed if the prediction error when the parallax vector is obtained is equal to or less than a certain threshold, and that the parallax compensation is not performed otherwise. In this case, the difference between the right-eye image 35 and the decoded image of the left-eye image 34 for which parallax has been compensated is obtained, and a difference image 36 is created.

The decision unit 103 performs block matching on the difference image 36 using the difference image 33. If the prediction error is equal to or less than a certain threshold value, motion compensation is performed. Otherwise, motion compensation is not performed. When it is determined that motion compensation is to be performed, a difference between the difference image 36 and the motion-compensated difference image 33 is obtained, and the difference value is encoded.

(3) If the decision section 103 decides not to perform motion compensation, the difference value at that time is encoded.

(4) When the determination unit 102 determines that the parallax compensation is not to be performed and the determination unit 104 determines that the flag MC is 1, the right-eye image 35 and the motion-compensated right-eye image 32 are compared. The difference from the decoded image is obtained, and the difference value is encoded.

(5) If the determination unit 104 determines that the flag MC is not 1, the pixel value of the right-eye image 35 is encoded as it is.

In (1) to (5), the difference image 36
Is not created, the difference between the decoded image of the left-eye image 34 and the right-eye image 35 that has been subjected to the disparity compensation using the disparity vectors obtained in (1) and (2) is obtained, and the difference image 3
Create 6.

That is, the image for the left eye is coded by motion compensation, and the image for the right eye is parallax compensation using the image for the left eye of the frame at the same time, motion compensation using the image for the right eye of the immediately preceding frame, or After parallax compensation using the left-eye image of the frame at the same time, encoding is performed by performing motion compensation using the difference image of the immediately preceding frame.

Thereafter, coding is performed by repeating the procedures (c) and (d). However, when coding by repeating the procedures (c) and (d), instead of the procedures (c) and (d), the coding according to the procedures (a) and (b) is performed in order to prevent the propagation of prediction errors. May be performed at regular intervals.

A method for encoding a left-eye image will be described with reference to FIG. In video coding, a method called motion compensation prediction coding using block matching, in which a coding unit is coded as a block (for example, composed of 16 × 16 pixels), is well known. . In the present embodiment, the coding method of the left-eye image is the same as the motion compensation prediction coding method using the block matching.

At this time, the image for the left eye is encoded for each frame. For example, as shown in FIG.
(N = 1, 2,...) Are input to the input unit 1, motion-compensated by the motion compensation unit 2 using the decoded image of the previous frame, and the detected motion vector is transmitted to the variable-length coding unit 6. Is done.

Thereafter, the difference between the left-eye image Ln and the motion-compensated image is calculated by the subtractor 3, and the difference data is converted and quantized through the conversion unit 4 and the quantization unit 5 to obtain a variable length code. Is transmitted to the conversion unit 6.

The difference data is sent to the variable length coding unit 6
, And through the inverse quantization unit 8 and the inverse transformation unit 9, the image subjected to inverse quantization and inverse transformation, and the motion compensated by the adder 10 is added, and the decoded image of the left-eye image Ln is stored in the frame memory. 11 is stored.

Next, a coding method of the right-eye image will be described. For example, as shown in FIG. 1, the right-eye image Rn (n =
1, 2,. . . ) Is input to the input unit 1.

The first image R1 for the right eye is compensated by block matching using the decoded image of the image L1 for the left eye. At this time, the right-eye image R1 and the decoded image of the left-eye image L1 from the frame memory 11 are input to the disparity compensation unit 12, and a disparity vector is obtained for each block.

The method for obtaining the disparity vector is as shown in FIG. 4 by superimposing an image 41 for the left eye and an image 42 for the right eye on the same plane as an image 43 and performing matching on the block of interest. Do. Therefore, one disparity vector is obtained for one block, and this disparity vector is obtained for all blocks.

The disparity vector is obtained by calculating the difference between the block of the image R1 for the right eye and the block extracted from the decoded image of the image L1 for the left eye, and has the smallest sum of the errors (hereinafter abbreviated as SAD) in the difference block. Choose a vector when At this time, the SAD is compared with the sum of the pixel values in the block, and if the SAD becomes smaller, it is determined that parallax compensation is to be performed.

In a block determined not to perform parallax compensation, a previously obtained parallax vector, if any, may be used instead of the parallax vector. For example, as shown in FIG. 5, in the block 52 determined not to perform the parallax compensation, when there is a block 51 to the left of the block 52, the block 52 may have the parallax vector of the block 51 instead of the parallax vector of the block 52. .

At this time, if the sum of the prediction errors due to the newly replaced disparity vector is equal to or larger than a certain threshold value, the previous disparity vector may be provided. These obtained disparity vectors are transmitted to the variable length coding unit 6 in FIG. 1 and are subjected to variable length coding. At this time, when a block is present on the left of the block to be coded, a difference from the disparity vector of the block may be obtained and transmitted to the variable length coding unit 6.

The right-eye image R1 and the decoded image of the left-eye image L1 having undergone parallax compensation are input to the subtractor 13, and the difference between them is obtained. The difference data is converted and quantized through a conversion unit 16 and a quantization unit 17, and is input to the variable length coding unit 6 and the inverse quantization unit 18, respectively. The variable length coding unit 6 performs variable length coding on the input data.

The difference data input to the inverse quantization unit 18 is inversely quantized, inversely transformed through the inverse transformation unit 19, added to the decoded image of the left-eye image L 1 having undergone parallax compensation by the adder 21, The decoded image of the right-eye image R1 is stored in the frame memory 22. Further, the data subjected to the variable length coding by the variable length coding unit is multiplexed by the multiplexing unit 7, respectively.

After that, the parallax compensating section 12 outputs the left-eye image L1.
, And performs parallax compensation on all the blocks using the decoded image and the disparity vector. A subtractor 23 creates a difference image between the compensated screen and the decoded image of the right-eye image R1. To store.

The second and subsequent right-eye images Rm (m = 2,
3,. . . ) Will be described. The encoding unit is a block, and the encoding is performed as follows. Right eye image R2
Is performed by the motion compensation unit 14 using the previous right-eye image R1 in the frame memory 22. If the SAD is smaller than the sum of the pixel values in the block, a motion vector MC is set to 1.

At this time, when the motion vector is (0, 0), it is determined that only the motion compensation is performed. Next, as in the case of the right-eye image R1 of the previous frame, the parallax compensator 12 performs block matching on the right-eye image R2 using the decoded image of the left-eye image L2 that is paired with the right-eye image R2 to obtain all disparity vectors. .

For example, as shown in FIG. 1, a right-eye image R2 is input to the input unit 1. The decoded image of the right-eye image R2 and the decoded image of the left-eye image L2 from the frame memory 11 are input to the disparity compensating unit 12 and block matching is performed to obtain a disparity vector.
And is variable-length coded.

When performing this block matching, when the motion compensation is first performed, the motion vector is (0, 0).
It is determined that parallax compensation is not to be performed for a block that is determined to perform only motion compensation in step S5, and the SAD of a block whose motion vector is (0,0) and that is not determined to perform only motion compensation is determined by the If the sum is smaller than the sum of the pixel values in, it is determined that parallax compensation is not to be performed, otherwise, it is not to be performed.

The decoded image of the right-eye image R2 and the parallax-compensated left-eye image L2 is input to the subtractor 13, and the difference between the right-eye image R2 and the decoded image of the parallax-compensated left-eye image L2 is obtained. The difference data is input to the motion compensation unit 14 and the subtracter 15, respectively.

The flag MC is set to 1 for a block of the data input to the motion compensating unit 14 for which the parallax compensating unit 12 determines not to perform the parallax compensation.
, The right-eye image R1 in the frame memory 22
, And motion compensation is performed using the previously obtained decoded image and the previously obtained motion vector.

Of the blocks of data input to the motion compensating unit 14, the blocks for which the parallax compensation has been determined to be performed and the parallax compensation has been performed are the same as the decoded image of the left-eye image L 1 in the frame memory 24. Motion compensation is performed using a difference image before the decoded image of the right-eye image R1.

As described above, when motion compensation is performed using the reference block extracted from the previous difference image, the reference block having the smallest SAD of the difference block is used for prediction. After obtaining the SAD, the variance may be obtained for each difference block, and the reference block having the smallest variance may be used for prediction.

The reference block having the smallest SAD among the reference blocks whose variance is equal to or less than a certain threshold may be used for prediction. In addition, the SAD and the variance may be ranked in order from the smallest one, weights may be given to both ranks, and the reference block when the sum of them is the smallest may be used for prediction.

Further, if there is an error value of a pixel unit in the difference block which is larger than a predetermined threshold value, the motion compensation may not be performed. Further, when compensating for a block having a block on the left side, the compensation method of the block on the left side is a method of performing only parallax compensation, a method of performing only motion compensation, and a motion compensation after performing parallax compensation. , A predetermined value may be subtracted from the value of the SAD by the compensation method so that the compensation method of the block on the left is easily selected.

After all blocks have been compensated for, in a block in which there are blocks near 8 surroundings, all blocks around 8 blocks around the block have the same compensation method, and Is different from the compensation method of the block near 8 surroundings,
The compensation method of the block may be made to match the compensation method of the blocks near eight surroundings.

At this time, a flag indicating that only the motion compensation is performed from the motion compensating unit 14, a flag indicating that only the parallax compensation is performed, or a flag indicating that the motion compensation is performed after the parallax compensation is performed and its motion vector Is transmitted to the variable-length coding unit 6 and is subjected to variable-length coding.

When the right-eye image R2 is input to the subtractor 15, the motion-compensated decoded image of the right-eye image R1 is input to the subtractor 15, and the difference between them is calculated. Further, the decoded image of the left-eye image L2 and the right-eye image R2
Is input to the subtractor 15, the decoded image of the right-eye image R1 subjected to motion compensation or motion compensation after parallax compensation is input to the subtractor 15, and the difference between them is obtained.

The difference data is converted by the conversion unit 16 and the quantization unit 1
7, the data is transformed and quantized, and input to the variable length coding unit 6 and the inverse quantization unit 18, respectively. The variable length coding unit 6 performs variable length coding on the input data.

The difference data input to the inverse quantization unit 18 is inversely quantized, inversely transformed through an inverse transformation unit 19, added to a motion-compensated decoded image of the right-eye image R 1 by an adder 20, and Further, the left-eye image L subjected to parallax compensation by the adder 21
2 and is stored in the frame memory 22 as a decoded image of the right-eye image R2.

The data variably coded by the variable length coding unit is multiplexed by the multiplexing unit 7. Thereafter, the parallax compensator 12 performs parallax compensation on all blocks using the decoded image of the left-eye image L2 and the parallax vector, and the subtractor 23 calculates the difference between the compensated image and the decoded image of the right-eye image R2. Create an image and store the data in frame memory 2
Store in 4. Hereinafter, the image for the right eye is encoded by this repetition.

In the present embodiment, the description has been given of the encoding of the binocular stereoscopic image. However, the encoding can be similarly performed for the multi-viewpoint stereoscopic image.

In the present embodiment, the description has been made of the method of performing only parallax compensation, the method of performing only motion compensation, and the method of performing motion compensation after performing parallax compensation for encoding the right image. A method of performing only parallax compensation, a method of performing only motion compensation, and a method of performing parallax compensation after performing motion compensation can be similarly used.

FIG. 6 shows an image decoding apparatus according to the present embodiment. The decoding device of the present embodiment is for decoding data encoded using the encoding device of FIG. The encoded data is input to the separation unit 61, and is separated into encoded data and a motion vector for the left eye image, encoded data and a motion vector and a disparity vector for the right eye image,
Each variable-length decoding unit 62 performs variable-length decoding.

The coded data of the image for the left eye is generally decoded in the same manner as the decoding method using motion compensation prediction, and the decoded image for the left eye is stored in the frame memory 63. The decoding of the encoded data of the first right-eye image will be described.

The encoded data of the first right-eye image is sent from the variable length decoding unit 62 to the inverse quantization unit 70 and the inverse transformation unit 71.
, And are inversely quantized and inversely transformed. The disparity compensation unit 74 receives the disparity vector from the variable length decoding unit 62 and the left-eye decoded image as a reference image from the frame memory 63, and creates a left-eye decoded image with parallax compensation using these.

At this time, the parallax compensating unit 74 creates a difference image obtained by calculating the difference between the decoded image for the left eye and the decoded image for the right eye for all the blocks, and stores the difference image in the frame memory 76. The adder 72 adds the parallax-compensated decoded image for the left eye to the encoded data of the image for the right eye to create a decoded image for the right eye.

The decoded image for the right eye is stored in the frame memory 75.
And output to the image display unit 78. The decoding of the encoded data of the second and subsequent right-eye images will be described.
The encoded data of the image for the right eye is inversely quantized from the variable length decoding unit 62 through the inverse quantization unit 70 and the inverse transformation unit 71, inversely transformed, and the difference data of the image for the right eye is decoded.

The motion compensator 77 uses the decoded image for the right eye in the frame memory 75, the difference image obtained by calculating the difference between the decoded image for the left eye and the decoded image for the right eye in the frame memory 76 as a reference image, and uses the motion vector to calculate the motion. Make compensation,
A first prediction image is created. The parallax compensating unit 74 performs parallax compensation on the motion-compensated first predicted image using a parallax vector to generate a second predicted image.

At this time, the parallax compensator 74 creates a difference image obtained by calculating the difference between the left-eye decoded image and the right-eye decoded image for all blocks, and stores the difference image in the frame memory 76. The first predicted image and the second predicted image are added to the decoded difference data of the right-eye image by an adder 72 and an adder 73, respectively, to generate a right-eye decoded image.

The right-eye decoded image is stored in the frame memory 7
5 and output to the image display unit 78. Thereafter, the encoded data of the right-eye image is decoded by this repetition.

In the present embodiment, a binocular stereoscopic image is used, but the same applies to a multi-viewpoint stereoscopic image.

[0076]

As described above, according to the coding apparatus of the present invention, only the parallax compensation is performed in the coding of the parallax image.
Although only three types of compensation methods are used, that is, motion compensation only and then motion compensation after parallax compensation, when only motion compensation is performed and the motion vector is (0, 0), only motion compensation is performed. In addition, it is possible to reduce the entire information amount of the vector required for encoding, and when there is a stationary portion such as a background in an image, matching errors are reduced, and the image quality is maintained while maintaining a high compression ratio. Can be improved.

According to the encoding apparatus of the present invention, in the encoding of a parallax image, when motion compensation is performed after performing parallax compensation or motion compensation or parallax compensation on a block having a block on the left. By subtracting a predetermined value from the SAD with the same compensation as that of the block on the left, block distortion can be reduced.

According to the encoding apparatus of the present invention, after compensating all blocks in the encoding of a parallax image, in a block where there are blocks near 8 surroundings, 8 blocks around the block exist. Are all of the same compensation scheme, and the compensation scheme of the block is different from the compensation scheme of the blocks in the neighborhood of eight, the block compensation scheme is adjusted to the compensation scheme of the blocks in the neighborhood of eight. Distortion can be reduced.

According to the encoding apparatus of the present invention, in encoding a parallax image, the variance is obtained for each difference block after obtaining the SADs of all the difference blocks, and the reference block having the smallest variance is used for prediction. Thereby, the image quality can be improved.

At this time, the image quality can be improved by using the reference block having the smallest SAD among the reference blocks whose variance is equal to or less than a certain threshold value for prediction.

At this time, the SAD and the variance are ranked in order from the smallest one, weights are given to both ranks, and the reference block when the sum of them is the smallest is used for prediction. Image quality can be improved.

At this time, if there is an error value in a pixel unit in the difference block after performing the motion compensation after performing the parallax compensation which is larger than a predetermined threshold, the motion compensation is not performed. Image quality can be improved.

Also, when encoding a parallax image, the compression ratio can be improved by performing parallax compensation after performing motion compensation instead of performing motion compensation after performing parallax compensation.

According to the encoding apparatus of the present invention, when obtaining a disparity vector for each block, the disparity vector of the block determined not to perform the disparity compensation is set as the first disparity vector, and the disparity compensation is performed before the block. When there is a block determined to be performed, a second parallax vector of the block determined to perform the parallax compensation before the block is used instead of the first parallax vector, and the second parallax vector is used. When the prediction error at the time is larger than a certain threshold, by having the first disparity vector,
The information amount of the disparity vector can be reduced.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of an encoding device that encodes a binocular stereoscopic image according to an embodiment of the present invention.

FIG. 2 is a diagram illustrating an order in which a stereoscopic image is encoded.

FIG. 3 is a diagram illustrating a coding method of a right-eye image.

FIG. 4 is a diagram illustrating an example of calculating a disparity vector.

FIG. 5 is a diagram for explaining that a block without parallax compensation has a parallax vector of a block on the left side.

FIG. 6 is a configuration diagram of a decoding device that decodes a binocular stereoscopic image.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Image input part 2, 14, 77 Motion compensation part 3, 13, 15, 23 Differentiator 4, 16 Transform part 5, 17 ... Quantization part 6 Variable length encoding part 7 Multiplexing part 8, 18, 70 Inverse quantum Transformation unit 9, 19, 71 Inverse transformation unit 10, 20, 21, 72, 73 Adder 11, 22, 24, 63, 75, 76 Frame memory 12, 74 Parallax compensation unit 30 Stereoscopic image 31, 34, 41 For left eye Image 32, 35, 42 Right-eye image 33, 36 Difference image 43 Synthetic image 51, 52 Block 61 Separation unit 62 Variable length decoding unit 78 Image display unit 100, 101, 102, 103, 104 Judgment unit

Claims (9)

[Claims] 1. A multi-viewpoint image encoding apparatus, wherein a first viewpoint image is determined as a reference image and motion compensation is performed on the reference image, and an image of a viewpoint other than the reference image is provided. A parallax compensating unit for performing parallax compensation on the parallax image, a second motion compensating unit for performing motion compensation on the parallax image or the parallax image subjected to parallax compensation, and a compensation method for each block. A multi-viewpoint image encoding apparatus, comprising: a determination unit for selecting, and when the second motion compensation unit determines that there is no motion amount, performs only motion compensation without performing parallax compensation. 2. A multi-view image encoding apparatus, wherein a first viewpoint image is determined as a reference image, and a first motion compensation means for performing motion compensation on the reference image, and an image of a viewpoint other than the reference image. A parallax compensating unit for performing parallax compensation on the parallax image, a second motion compensating unit for performing motion compensation on the parallax image or the parallax image subjected to parallax compensation, and a compensation method for each block. And a determination unit for selecting, in the encoding of the parallax image, for a block in which a block is present on the left side, parallax compensation,
Alternatively, when performing motion compensation after performing motion compensation or parallax compensation, a predetermined value is subtracted from the sum of errors in a difference block by the same compensation as the block on the left. Multi-view image coding device. 3. A multi-view image coding apparatus, wherein a first viewpoint image is determined as a reference image, and motion compensation is performed on the reference image, and an image of a viewpoint other than the reference image is provided. A parallax compensating unit for performing parallax compensation on the parallax image, a second motion compensating unit for performing motion compensation on the parallax image or the parallax image subjected to parallax compensation, and a compensation method for each block. A determination unit for selecting, and in the parallax image coding, after compensating for all blocks, in a block in which blocks around 8 exist, all blocks around 8 around the block If the same compensation is performed on the block and the compensation performed on the block is different from the compensation on the blocks near the surroundings 8, the compensation performed on the block is A multi-viewpoint image encoding apparatus, which is adapted to compensate for blocks near eight. 4. A multi-viewpoint image encoding apparatus, wherein a first viewpoint image is determined as a reference image and motion compensation is performed on the reference image, and an image of a viewpoint other than the reference image is provided. A parallax compensating unit for performing parallax compensation on the parallax image, a second motion compensating unit for performing motion compensation on the parallax image or the parallax image subjected to parallax compensation, and a compensation method for each block. And a determination unit for selecting, in the encoding of the disparity image, when performing motion compensation after performing disparity compensation, determine the prediction error and variance in the block, and predict the reference block when the variance is minimized. A multi-view image encoding device characterized by using: 5. The multi-view image encoding apparatus according to claim 1, wherein, when encoding the parallax image, motion compensation is performed instead of performing motion compensation after performing parallax compensation. A multi-viewpoint image encoding apparatus, which performs parallax compensation after the image processing. 6. The multi-view image encoding apparatus according to claim 4, wherein, among the reference blocks whose variance is equal to or less than a certain threshold, a reference block having the smallest sum of prediction errors in the block is used for prediction. A multi-view image encoding device characterized by the following. 7. The multi-view image encoding apparatus according to claim 4, wherein the sum total value and the variance of the errors in the block are ranked in order from the smallest one, and both ranks are weighted. A multi-view image coding apparatus characterized in that a reference block when a value obtained by adding them is the smallest is used for prediction. 8. The multi-view image encoding apparatus according to claim 1, wherein a prediction error when performing motion compensation after performing parallax compensation is larger than a predetermined threshold value when viewed in pixel units. A multi-view image coding apparatus, wherein motion compensation is not performed if there is any. 9. The multi-view image encoding apparatus according to claim 1, wherein when obtaining a disparity vector for each block, a disparity vector of a block determined not to perform disparity compensation is determined by a first disparity vector. When there is a block determined to perform disparity compensation before the block as a vector, a second disparity vector of the block determined to perform disparity compensation before the block is used instead of the first disparity vector. A multi-view image encoding apparatus having a first disparity vector when a prediction error when the second disparity vector is used is larger than a certain threshold.