JPH01114283A - Method for compressing and reproducing redundancy of stereoscopic picture - Google Patents

Abstract

PURPOSE:To compress redundancy by setting one picture to reference among plural pictures, dividing others into picture blocks based on the contour lines of an object, shifting them in parallel in a horizontal direction and executing a differential processing with a reference picture in the position. CONSTITUTION:The picture is divided into strips in the horizontal directions and respective strip pictures are divided in vertical directions based on the characteristic of the pictures. Respective divided pictures obtained and the reference picture are executed alignment of position, and the differential processing is executed in the position. Consequently, the object in a different depth in correspondence with the overlap and separation of the object can be separated without increasing the number of the blocks in the whole picture even if the objects have infinite volumes. Thus, redundancy can be compressed.

Description

[Detailed Description of the Invention] [Industrial Application Field] This invention relates to a redundancy compression method and reproduction method for stereoscopic images, and in particular, to displaying a face image of the other party as a stereoscopic image in a system such as a videophone or a video conference. The present invention relates to a stereoscopic image redundancy compression method and a reproduction method in which redundancy is compressed when transmitting a stereoscopic image and reproduced on a receiving side.

[Prior Art] FIG. 11 is a schematic block diagram showing the configuration of a conventional three-dimensional video conference system, and FIGS.
FIG. 4 is a diagram for explaining a model of a stereoscopic image capturing system.

In FIG. 11, an image of the conference attendee 1 on the transmitting side is simultaneously captured from two directions by a right eye camera 2a corresponding to the right eye and a left eye camera 2b corresponding to the left eye. These right eye camera 2a and left eye camera 2b are each constituted by a television camera. Right eye camera 2a, left eye camera 2
The image signals that are the shooting outputs of b are sent to the transmission control device 3.
a, 3b, and transmitted to reception control devices 5a, 5b via transmission lines 4a, 4b. Reception control device 5a, 5
b causes the stereoscopic display device 8 to display images corresponding to the images seen from the left and right eyes, respectively.

In addition, in FIG. 11, in order to simplify the illustration,
Only one-way communication is shown; of course, bi-directional communication is also possible, and the audio transmission circuit is also omitted.

The model of the three-dimensional image capturing system will be explained in more detail with reference to FIGS. 12, 13, and 14. The right eye camera 2a includes a right eye lens 23a, and this right eye lens 23
The right eye camera focal point 24a is located on the optical axis 25a'' of ``a''. Similarly, the left eye camera 2b includes a left eye lens 23b, and the left eye camera focal point 24b is located on the optical axis 25b of the left eye lens 23b. An object in the field of view of the right eye camera 2a is imaged on a negative image plane 26a by the right eye lens 23a, and an object in the field of view of the left eye camera 2b is imaged by the left lens 23a.
b, the image is formed on the negative image plane 26b.

The images on the negative image planes 26a and 26b are both inverted images with the right and left sides reversed, and are converted into erect images by the transmission control devices 3a and 3b shown in FIG. 11. This erect statue is
Geometrically, the positive image surfaces 27a and 27b shown in FIG.
The images on the positive image planes 27a and 27b become the right-eye image and the left-eye image, respectively, and are displayed on the stereoscopic display device 8 shown in FIG. 11 to reproduce the stereoscopic image.

As shown in FIG. 13, when objects A, B, and C exist in the field of view of the right-eye camera 2a and the left-eye camera 2b in the depth direction, the right-eye positive image plane 27a and the left-eye positive image plane The order in which the objects A, B, and C photographed in 27b are reversed.

In FIG. 13, the volumes of objects A, B, and C are schematically shown as being infinitesimal. FIG. 14 shows a case where the volume of the object has a finite size in order to bring it closer to reality.

As shown in FIG. 14, when objects 9 and 10 exist in the field of view of the right eye camera 2a and the left eye camera 2b in the depth direction, they are reflected on the right eye positive image plane 27a and the left eye positive image plane 27b. The images of the objects 9 and 10 overlap or separate depending on the positions of the right eye camera 2a and the left eye camera 2b.

By presenting these images to the human's right and left eyes, respectively, it is possible to perceive that the object is located at the appropriate position in the depth direction.

Based on the above-mentioned principle, the stereoscopic display device 8 shown in FIG. The expression allows images to be displayed with a sense of realism.

However, in the video conference system shown in FIG. 11, it is necessary to transmit twice as much image information as seen from the left and right eyes to the receiving side via the transmission paths 4a and 4b. For this reason, there is a problem in that it puts pressure on the financial burden of line usage on users. Furthermore, as the stereoscopic display device 8, for example, there is a lenticular plate system, but in order to display a natural stereoscopic image using such a lenticular plate system, images viewed from about five directions are also required, which requires an enormous amount of processing time. The problem is that the amount of information is large, and it is impractical to transmit this amount.

In addition, when considering the storage of shooting records of stereoscopic images, rather than telecommunications, it has the disadvantage that it consumes twice as much storage medium (memory) as conventional monocular images.

For this reason, there were also problems in that securing a storage space and the cost of recording media became high.

DPC is a method for compressing such a huge amount of information.
A method is known in which M is applied not in the time direction but between the left and right image frames. Such a method is described by Lukac
s M, Michael; Predictive
Coding orMulti-viewpoint
image 5ets'', Proced]ng of
JEEE, Int, Conr.

Acoust, 5peech Signal
Proccss,, Vol. 986° No.
1. P521-524.1986. Lukacs' method divides each frame of a monocular image into N blocks each consisting of XXY pixels, spatially moves paired image blocks in the horizontal direction, and performs differential processing between images to compress redundancy. are doing.

[Problems to be Solved by the Invention] However, in Lukacs' method, the block size (1 block = XxY pixels) cannot be made variable within one image according to the characteristics of the image. It had the following drawbacks and could not be put to practical use.

That is, when the object as shown in FIG. 14 has a finite volume, the images overlap by 1M. For this reason, conventional uniform block division cannot separate objects at different depths, and the redundancy of the differential image cannot be reduced to the desired degree, resulting in poor redundancy and compression efficiency.

In addition, in order to better separate different objects using conventional methods, it is possible to reduce the block size, but when the block size is reduced, the number of blocks in the entire image increases, and each block is As a result, the amount of codes increases, resulting in poor redundancy and compression efficiency.

Furthermore, in the method of reducing the block size described above,
As the number of blocks increases, the number of calculations for moving each block increases, and in order to shorten the calculation time, it is necessary to increase the size of the processing circuit, resulting in the problem that the device becomes expensive.

Therefore, the main object of the present invention is to provide a method for compressing the redundancy of a stereoscopic image and a method for reproducing the redundancy of the stereoscopic image, which can satisfactorily compress the redundancy of a stereoscopic image, transmit or store it, and reproduce it.

[Means for Solving the Problem] In the stereoscopic image redundancy compression method and reproduction method according to the present invention, an image photographed from at least one direction among images photographed from a plurality of directions is used as a base image; The image taken from the direction is divided horizontally into strips, each strip image is divided based on the image characteristics, and each divided image is translated in the horizontal direction so that it overlaps with the reference image. The difference between the reference image and the divided image is calculated for each corresponding pixel at the moved position, and a three-dimensional image is expressed based on the obtained multiple difference images, both parallel images, and the reference image, thereby reducing redundancy. It is configured to compress.

In addition, when reproducing a stereoscopic image based on an image whose redundancy has been compressed in this way, multiple difference images and a reference image are added, and these added images are applied to the translation vector for each. Then, the original image is reproduced by parallel movement in the opposite direction, and a stereoscopic image is reproduced based on the reproduced image and the reference image.

[Operation] In the three-dimensional image redundancy compression method and reproduction method according to the present invention, an image is divided horizontally into strips, each strip image is divided vertically based on image characteristics, and each of the obtained Since the divided images and the reference image are aligned and the difference processing is performed at that position, there is no need to increase the number of blocks in the entire image, and even when the object has a finite volume. It is also possible to separate objects at different depths depending on whether objects overlap or separate.
Redundancy can be effectively compressed by differential processing.

Furthermore, the original image can be easily reproduced by simply adding the difference image and the reference image and translating the added image in the opposite direction to the translation vector for each, and based on the reproduced image and the reference image. 3D images can be played back.

In the conventional method, the entire input image is divided into blocks of the same shape, whereas the present invention differs in that it can be divided according to the complexity of the input image.

[Embodiments of the Invention First Embodiment FIG. 1 is a schematic block diagram showing the overall configuration of an embodiment of the present invention, and FIG. 2 shows a more specific diagram of the division processing section shown in FIG. 1. It is a block diagram.

First, with reference to FIG. 1, the overall configuration of a first embodiment of the present invention will be described. The image signals output from the right eye camera 2a and the left camera 2b are sent to an A/D converter 28a.
, 28b, the signals are sequentially converted into digital signals in units of pixels. The right eye image signal output from the right eye camera 2a is divided into four rectangular images by the image horizontal divider 29. The four rectangular images are input terminals 56
The signal is given to each division processing unit 42 via. Division processing unit 4
2 will be explained in detail in FIG. 2, which will be described later.

An image signal output from the left eye camera 2b and converted into a digital signal by the A/D converter 28b is input to the division processing section 42 via the input terminal 57. The division processing unit 42 compresses the redundancy of the strip-shaped image inputted from the input terminal 56 based on the image signal inputted manually to the input terminal 57, and outputs the compressed image data to the output terminal 58.
Output to. The compressed image data output from the output terminal 58 is given to the modulator 39a.

The modulator 39a converts the compressed image data into a signal format and transmission rate suitable for the transmission path 4a, and sends the converted image data to the transmission path 4a. On the other hand, the left eye image signal output from the left eye camera 2b is directly given to the modulator 39b, converted into a signal format and transmission rate suitable for the transmission path 4b, and sent out to the transmission path 4b.

Next, the configuration of the division processing section 42 will be explained with reference to FIG. The rectangular image signal input to the division processing section 42 is applied to an edge detector 30, and the outline of the image is extracted. The extracted contour image signal is sent to a divided vertical line generator 4.
given to 3. The dividing vertical line generator 43 generates a vertical dividing line for vertically dividing the rectangular image so as to intersect the contour line based on the contour line image signal, and generates a vertical dividing line for vertically dividing the rectangular image so as to intersect the contour line.
Give to 1. The image vertical divider 31- further divides the rectangular image in the vertical direction based on the vertical line data generated by the dividing vertical line generator 43. Since the contour lines in the image almost coincide with the boundaries of each object, it is possible to obtain image blocks in which the objects are separated by depth. Thereby, the effects of this invention can be exhibited. The image blocks are stored in the memory 40 and input to, for example, three vertical division processing units 41, where they are sequentially distributed and processed.

In the vertical division processing section 41 , the input image block is translated in the horizontal direction by the translation processor 33 by the same amount as three different translation vectors generated from the translation vector generator 32 . The image signals that have been translated in parallel by three vector amounts by the translation processor 33 are processed by differentiators 341, 342, and 343, respectively.
given to. Differentiators 341, 342 and 343 are
The difference component between the image signal output from the left eye camera 2b and converted into a digital signal by the A/D converter 28b and the right eye image signal that has been translated in parallel is calculated for each pixel. And differentiators 341, 342 and 343
The respective outputs will be explained in detail in FIGS. 3A and 3B below, but are given to difference region extractors 351, 352, and 353, and only the regions differentiated by the image blocks from the difference images are extracted. and is given to image information amount calculators 361, 362 and 363.

Image information amount calculator 361. ．． In 362 and 363,
The cumulative residual of the input difference image block is calculated.

Note that in the configuration shown in FIG. 2 described above, to simplify the explanation, image translation processing, difference processing, and cumulative residual calculation processing are performed based on three parallel translations. Of course, this is not limited to three parallel displacements, and it goes without saying that by performing the above-described processing on more than three parallel displacements, the redundancy of the image can be compressed more precisely and effectively.

” The residuals calculated by the image information amount calculators 361, 362, and 363 in two series are fed to the comparator 37, where the residuals are compared, and the smallest residual is outputted and given to the selector 38. It will be done. The selector 38 is given difference image blocks from the difference region extractors 351, 352, and 353, and the selector 38 selects the parallel translation and difference image blocks corresponding to the minimum residual given from the comparator 37. is applied to the modulator 39a.

The modulator 39a converts the parallel movement and differential image blowers into a signal format and transmission rate suitable for the transmission line 4a, and sends the signals to the transmission line 4a.

Next, the specific operation of the first embodiment of the present invention will be described with reference to FIGS. 1, 2, 3A, and 3B. In step 1 shown in FIG. 3A, the right eye image signal photographed by the right eye camera 2a is sent to the A/D converter 28a.
is converted into a digital signal by This right eye image is represented by I r (x, y). Further, the left eye image signal photographed by the left eye camera 2b is converted into a digital signal by the A/D converter 28b, and this left eye image is represented by Iu (x, y). In step 2, the right eye image 1 r (x, y) is horizontally divided into four rectangular images Ir+ by the image horizontal divider 29.
(x, y), Ir2 (x, y), Ir3 (x.

y), Ir4 (x, y).

Next, in step 3, the edge detector 30 extracts the contour line of the rectangular image, and the dividing vertical line generator 43 extracts the contour line of the rectangular image.
given to. The dividing vertical line generator 43 calculates a vertical line to further divide the image in the vertical direction based on the given contour line and supplies it to the image vertical divider 31 . The vertical image divider 3 divides the image into, for example, six image blocks “I l+ ”+2+ Ir+ 3+ with the vertical line as the boundary.
Divided into ``+4+I'15+''+6 memory 40
Accumulates in. The other division processing units 42 similarly perform sequential parallel processing.

FIG. 4 shows a specific example of the edge detector 30 shown in FIG. In the example shown in FIG. 4, for each pixel, the difference between the left and second neighboring values (i.e., I(x-1, y) and I(x, y-1j) is 1') for the input image. (x, y) (However, 1' (x, y)
=l I (x, y) -I (x~1゜y)
l+l I (x, y) -I (x, y-1,
) By calculating l), the boundary line of the object appearing in the input image can be output.

Here, edge detection is performed based on the difference between adjacent pixels; however, the present invention is not limited to this, and other detection methods may be used as long as the outline of the object can be calculated. Needless to say, the same effect can be obtained. In addition, a zero-crossing method that calculates second-order differences is also effective.

FIG. 5 specifically shows the operation of the divided vertical line generator 43. The divided vertical line generator 43 is connected to the edge detector 30
Based on the output image I' r, (x, y) from
The average value of the X coordinates of the detected edges is calculated and provided to the image vertical divider 31 as a dividing vertical line. As shown in FIG. 5, the image vertical divider 31 divides the rectangular image Ir, (x, y) into two image blocks Ir, I (x, y), Ir,
Subdivide into 2(x, y).

FIG. 6 is a diagram for explaining the parallel processing operation by the parallel movement processor 33. In step 4, the parallel shift processor 33 translates the divided image blocks in units of pixels, as shown in FIG. for example,
Image blocks Ir, divided by the image vertical divider 31.

(x, y), three translation vectors d, d2 . Ir++(x1+y+), Ir++(x1+y+),
■r, I (x2.V2)l Ir+ +(X3.
! /a).

In step 5 shown in FIG. 3B, the differentiator 341.3
Left eye image IQ (x, y) by 42 and 343
The difference is calculated for each corresponding pixel. Next, in step 6, the difference region extractors 351, 352 and 3
53, only the difference regions are extracted, and the difference image blocks A,,B,.

C1 is calculated. Furthermore, in step 7, the cumulative residual of each block is determined according to the following arithmetic expression.

Next, in step 8, each cumulative residual is compared by the comparator 37, and in step 9, the translated and difference image blocks giving the minimum cumulative residual are selected by the selector 38 and applied to the modulator 39a. .

As described above, according to the first embodiment of the present invention, even in actual stereoscopic images in which objects at different depths in the left and right images shown in FIG. The left image of the object can be distributed to each block, and the common part between the left and right images can be accurately removed by differential processing by parallel translation processing of each block, and the common part is not transmitted, which is faster than conventional technology. The amount of information can be significantly compressed.

Note that step 2 shown in FIGS. 1 and 3A above
In the explanation, in order to simplify the explanation, the A/D converter 2
Although the right eye image inputted from 8a is divided into four parts in the horizontal direction, and each part is processed by the corresponding division processing section 42, the present invention is not limited to this. It goes without saying that by dividing the image into two in the horizontal direction and processing the image using the division processing unit 42, the common portion between the left and right images can be accurately removed, and the redundancy of the image can be effectively compressed. To achieve this, it is possible to increase the number of division processing units 42 shown in FIG. This can also be realized by providing a memory for temporarily storing divided pixels between the unit 42 and sequentially inputting divided images from the memory to the division processing unit 42 for processing.

In addition, in FIG. 1, for simplification of explanation, regarding stereoscopic images input from two television cameras, one is assumed to be a right-eye image and the other is assumed to be a left-eye image, and the right-eye image is compared to the left-eye image. However, the present invention is not limited to such two-lens input images, but also applies to two or more multi-lens images, with at least one of them being the reference image, and the other images being compressed. A/D converter 28a1 image horizontal divider 29 shown in FIG.
By providing the division processing unit 42 and the modulator 39a, redundancy can be compressed.

Second Embodiment FIG. 7 is a concrete block diagram of the vertical division processing section 41 in a second embodiment of the present invention. The embodiment shown in FIG. 7 is a modification of the vertical division processing section 41 shown in FIG. 2 described above. In this second embodiment, the right eye and left eye image signals are color image signals consisting of R, G, and B colors, and the A/D converter shown in FIG. 28a, a plurality of image block signals are generated via two image horizontal dividers and an image vertical divider 31, and one of the image block signals is input to the vertical division processing unit 4. It is applied to terminal 53.

Similarly, the left eye image signal is also a color image signal consisting of R, G, and B colors, and the A/D converter 2 shown in FIG.
8b and input to the input terminal 54.

An image block signal composed of each color of R, G, and B input to the input terminal 53 is applied to an information reduction filter 441. The information reduction filter 441 selects only the R image component from the image block signal, and further performs thinning processing to remove, for example, every other pixel out of -22= pixels in one screen, and reduce the resolution to 1.
/2, reducing the amount of information of the right eye image signal. Further, the left eye image signal input to the input terminal 54 is provided to an information reduction filter 442, where R image selection and thinning processing are performed.

In this way, for the quantized left and right R image signals, the parallel displacement generators 32
The comparator 37 performs parallel movement processing, difference processing, cumulative residual calculation processing, and comparison processing, and outputs the parallel movement that provides the minimum cumulative residual. Then, based on the translation, the unquantized right eye image block is translated by the image translation processor 331, and the difference with the left eye image is calculated by the subtractor 344. Further, the differential area extractor 354 outputs an image from which the differential area has been extracted.

According to the second embodiment of the present invention, parallel movement processing is performed on an image with a reduced amount of information.

Since the cumulative residual calculation process and the cumulative residual comparison process are performed, it is possible to effectively compress the amount of information of a stereoscopic image while reducing the circuit scale for performing these processes and increasing the speed of calculation.

In the embodiment shown in FIG. 7, the information reduction filter 441 selects the R image among the R, G, and B image signals, but the invention is of course limited to this. Instead, the amount of information in the processed image can be reduced by selecting the G component or the B component, or by reducing the number of gradation levels, effectively compressing the amount of information in the 3D image. can be achieved.

Third Embodiment FIG. 8 is a block diagram of a third embodiment of the present invention, and FIG. 9 is a diagram for explaining an example of dividing a right eye image in the third embodiment.

In FIG. 8, the vertical division processing section 41 is almost the same as the second embodiment except for the following points. That is, the information amount reduction filter 443 is provided outside the vertical division processing section 41, and the information amount reduction filter 443 reduces the information amount of the left eye signal from the D/A converter 28b. moreover,
In this third implementation=24- example, even when the right eye image is horizontally divided, it is divided based on the characteristics of the image. The operation will be explained in detail below.

In this third embodiment, similarly to the second embodiment,
The right eye and left eye image signals are color image signals consisting of each color of R, G, and B. These color image signals are converted into digital signals by the A/D converter 28a, and the edge detector 30 detects the outline of the image. is extracted. The contour line image signal is applied to a divided horizontal line generator 100, which generates divided horizontal line signals of the input image in the horizontal direction so as to intersect the contour line. The right eye image signal is input to the image horizontal direction divider 29, and when a divided horizontal line signal is given from the divided horizontal line generator 100, a rectangular image signal having an adaptive width is outputted.

In the first embodiment described above, as shown in step 2 of FIG. 3A, the rectangular image signals 1r, having the same width.

Ir2. Ir3. In this example, the input image is divided into multiple pairs according to the position of the object, with the input image being finely divided in places where the object is located, and wide in places where the object is not present. For example, as shown in FIG. 9, rectangular images 1r, , Ir2+ Ir3...
・It is divided into The divided rectangular image signals are stored in memory 4.
01, and input to, for example, two divided image processing units 42, where they are sequentially processed in parallel.

In the divided image processing unit 42, the edge detector 302 or the difference region extractor 354 shown in FIG. A difference image block that performs reduction processing, parallel movement processing, second difference processing, difference region extraction processing, cumulative residual comparison processing, and image selection processing and provides a difference cumulative residual is input to the modulator 39a. On the other hand, the left eye image is input as is to the modulator 39b. Modulator 39a, 3
9b converts the differential image block signal and the left eye signal into signal formats and transmission speeds that conform to the specifications of the transmission lines 4a and 4b, and sends them to the transmission lines 4a and 4b.

As described above, according to the third embodiment of the present invention, even when the left image is horizontally divided, a rectangular image is adaptively generated based on the characteristics of the input image, as shown in FIG. 9, for example. Therefore, when the subsequent vertical divisions are combined, objects at different depths can be divided into the first and second
Even better separation can be achieved than in the examples.

Furthermore, since the size of the image block can be increased in an image area where no target object is placed, it is possible to prevent an increase in the number of blocks in the entire image. In this way, the two-point series can achieve effective redundancy compression of stereoscopic images compared to the prior art.

Fourth Embodiment FIG. 10 is a schematic block diagram of a fourth embodiment of the present invention.

In FIG. 10, the transmitting section is constructed in the same manner as in FIG. 8 described above, so a detailed explanation thereof will be omitted, and only the receiving section will be explained. The parallel movement and difference image blocks are transmitted from the transmitter to the receiver via transmission paths 4a and 4b. The demodulator 45a demodulates the parallel movement and difference image blocks transmitted from the modulator 39a of the transmitter, and the demodulator 45b demodulates the left eye image signal transmitted from the modulators 3b. The demodulated parallel movement and difference image blocks are divided into four by the information divider 46, stored in memories 403 and 404, and sequentially input to the difference image processing unit 50 where they are processed in parallel.

In the difference image processing unit 50, the input parallel movement blocks are given to four inverse vector generators, and the difference image block is sent to an adder 4.
Give to 7. The adder 47 adds the difference image block and the demodulated left eye image signal and provides the result to the addition area extractor 48 . The addition region extractor 48 extracts only the added region and provides it to the translation processor 332.

On the other hand, the inverse vector I/L generator 49 outputs parallel movements in the opposite direction based on the input parallel movements and provides them to the parallel movement processor 332 . The parallel shift processor 332 performs a backward parallel shift process on the addition region extracted by the addition region extractor 48 based on the reverse parallel motions given from the inverse vector generator 49 . And parallel movement processor 3
The image signal translated by 32 is sent to an image synthesizer 51
and is combined with image signals processed by other differential image processing units 50. Then, the combined image signal is converted into an analog signal by the D/A converter 52a and provided to the stereoscopic display device 8.

Further, the left-eye image signal demodulated by the demodulator 45b is converted into an analog signal by the D/A converter 52b and provided to the stereoscopic display device 8.

Thereby, the stereoscopic image on the transmitting side is faithfully displayed on the stereoscopic display device 8. In this way, the redundancy of the stereoscopic image can be effectively compressed on the transmitting side, and the original images (left-eye image and right-eye image) can be faithfully reproduced from the compressed image on the receiving side. Images can be displayed on the stereoscopic display device 8.

In addition, in the above-mentioned third embodiment, in order to simplify the explanation,
The redundancy of the left eye image itself is not compressed and is sent as is to the receiving unit, but the A/D converter 28
With respect to the left eye image signal output from b, after the signal is input to the divided image processing section 42 and before being input to the modulator 39b, the left eye image signal is given to a redundancy compression section to reduce redundancy. The demodulator 45b and the D/A converter 5
A left eye image signal reproducing unit is provided between the transmission line 4b and the left eye image signal, and by reproducing the compressed left eye image signal, the information amount of the image transmitted to the transmission path 4b is appropriately compressed, and the image is faithfully reproduced on the receiving side. Needless to say, the signal can be reproduced.

As a result, the redundancy of the stereoscopic image can be comprehensively compressed and reproduced in conjunction with the already-described redundancy compression and reproduction of the right-eye image and the left-eye image.

Furthermore, the aforementioned left-eye image signal redundancy compression section and reproduction section employ methods conventionally used for compression and reproduction of television signals, such as DPCM (C,C,Cut le
r, "Dir"l'ercntjal Quanti
zation of Communication
Signals''.

U.S. Patent No. 2605361.
J IJ 1 y 2Q, 1.952).

Note that in the above embodiments, only those that transmit images in one direction have been described, but by adding components with the same configuration, reception and transmission can be performed at the same time, and bidirectional communication can be achieved. It goes without saying that this can be achieved.

In addition, although the above-mentioned embodiment shows the configuration for communication, by adding a recording medium and its control device as a component instead of the transmission paths 4g and 4b,
It becomes possible to effectively accumulate three-dimensional images.

[Effects of the Invention] As described above, according to the present invention, it is possible to focus on the fact that in images seen from multiple directions, the images are very similar even though they overlap or are separated from the object seen from each direction, At least one image among the multiple images is used as a brush image, the other images are each divided into image blocks based on the outline of the object, and each image block is horizontally shifted in parallel to the reference image. By performing differential processing, common parts between multiple images can be effectively removed and redundancy can be compressed. Furthermore, based on the redundancy-compressed difference image block, parallel translation, and reference image, a summation image of the difference image block and the reference image is calculated, and the summation image is parallelized in the opposite direction to the translation at the time of compression. By moving the image and reproducing a stereoscopic image based on the reproduced image and the reference image, it is possible to reproduce a stereoscopic image with excellent realism.

[Brief explanation of the drawing]

FIG. 1 is a schematic block diagram showing the overall configuration of a first embodiment of the present invention. FIG. 2 is a concrete block diagram of the division processing section shown in FIG. 1. FIGS. 3A and 3B are diagrams for explaining the operation of the first embodiment of the present invention. FIG. 4 is a diagram for explaining the operation of the edge detector shown in FIG. 2. FIG. 5 is a diagram for explaining the operation of the vertical image divider shown in FIG. 2. FIG. 6 is a diagram for explaining the operation of the parallel movement processor shown in FIG. 2. FIG. 7 is a schematic block diagram of a second embodiment of the invention. FIG. 8 is a schematic block diagram of a third embodiment of the present invention. FIG. 9 is a diagram for explaining an example of dividing the right eye image in the third embodiment. Figure 10 is the fourth example of this invention.
FIG. 2 is a schematic block diagram of an embodiment. FIG. 11 is a block diagram showing the general configuration of a conventional three-dimensional video conference. 1st
2, FIG. 13, and FIG. 14 are diagrams for explaining a model of a conventional stereoscopic image photographing system. In the figure, 2a is a right eye camera, 2b is a left eye camera, 4
a, 4b are transmission paths, 8 is a stereoscopic display device, 28a, 28b
is an A/D converter, 29 is an image horizontal divider, 30 is an edge detector, 31 is an image vertical divider, 32 is a parallel displacement generator, 33° 331.332 is a parallel displacement processor,
341, 342. 343, 344 are differentiators, 351, 3
52゜353.354 is a difference region extractor, 361, 36
2. 363 is an image information amount calculator, 37 is a comparator, 38 is a selector, 39a, 39b are modulators, 40°401 to 404 are memories, 4] is a vertical division processing section, 42 is a division processing section, 43 is Divided vertical line generation = 33- device, 441 to 445 are information reduction filters, 45a,
45b is a demodulator, 46 is an information divider, 47 is an adder, 4
8 is an addition area extractor, 4 is an inverse vector generator, 50 is a difference image processing unit, 51 is an image synthesizer, 52a, 52b (
, tD/A converter, 1-00 indicates a divided horizontal line generator. Patent applicant: ATR Tsushin Co., Ltd. Figure 5 Figure 6

Claims (5)

[Claims] (1) In a device that obtains a three-dimensional image based on images of the same object taken from multiple directions, dividing an image taken from at least one direction among the images taken from the plurality of directions into strips in the horizontal direction; Each strip image is divided vertically according to the image characteristics, each divided image is moved parallel to the reference image in the horizontal direction, and each pixel corresponding to the moved position is divided into the reference image. A redundancy compression method for a three-dimensional image, characterized in that the three-dimensional image is represented by a plurality of difference images, parallel images, and a reference image in which differences between the image and each divided image are determined. (2) Furthermore, each of the divided images is translated in parallel based on a plurality of translation amounts, and the difference between each translated divided image and the reference image is determined, and the amount of information in the difference image is calculated. 2. The redundancy compression method of a stereoscopic image according to claim 1, wherein the three-dimensional image redundancy compression method is characterized in that the three-dimensional images are compared and the parallel movement is performed based on the amount of parallel movement that provides the minimum amount of information. (3) The three-dimensional image according to claim 1, further comprising comparing the amount of information of images whose resolution or the number of tones and gradations has been reduced before the parallel movement process. redundancy compression method. (4) The stereoscopic image redundancy compression method according to claim 1, further comprising dividing the image horizontally into strips based on the characteristics of the image. (5) dividing a reference image photographed from at least one direction of the same object and images photographed from other directions into a plurality of images, and translating each divided image in the horizontal direction;
A stereoscopic image in which the difference between the reference image and each divided image is determined for each corresponding pixel at the movement position, and a stereoscopic image is reproduced based on the obtained plurality of difference images, the amount of parallel movement, and the reference image. In the reproduction method, the original image is reproduced by adding the plurality of difference images and the reference image, and translating the added images in the opposite direction with respect to the translation vectors for each. A method for reproducing a stereoscopic image, characterized in that a stereoscopic image is reproduced based on the reproduced image and the reference image.