JPH06217278A - Image data encoding device - Google Patents

Abstract

PURPOSE:To improve the efficiency of encoding by providing a parameter calcu lation part which supplies a weighting pattern corresponding to a current block to a weighting process part or forcibly sets the value of input image data on the side of a high frequency component to '0'. CONSTITUTION:Longitudinal and lateral position determination parts 3 and 4 detect the current longitudinal and lateral positions (x) and (y) of respective blocks from a horizontal and a vertical synchronizing signal, a clock signal, etc., by using a counter, etc. Then a parameter calculation part 5 selects the weighting pattern at the weighting process part 2 as preprocessing for quantization or encoding by using parameters which are given by an operator and indicate which position on a screen is increased in picture quality and the frequency components after orthogonal conversion by an orthogonal conversion part 1 are weighted. In another way, the values of more low frequency components of the input image data are set to '0' toward the peripheral part except in a high-picture-quality area.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image data encoding apparatus, and more particularly to an apparatus for encoding multi-view stereoscopic image data or three-dimensional volume data with high efficiency.

FIG. 12 shows a conceptual configuration of a system that has already been made public at the NHK Technical Research Institute in 1992 in the case of producing a multi-view stereoscopic image. First, a stationary or moving subject (octopus) 100 is Images are taken with a plurality of camera groups 110 whose positions are slightly shifted vertically and horizontally. Next, the image data obtained from the plurality of camera groups 110 is converted into the encoder 1
High-efficiency coding is performed at 20, multiplex is performed at the multiplexing unit 130, and then transmitted via the transmission path 140 and the like. At the receiving side, demultiplexing is performed at the separating unit 140, decoding is performed, and the result is displayed on the display 160.

As the display 160 on the output side, for example, a lenticular lens (when there is parallax only in the horizontal direction) or a fly's eye lens (when there is parallax in the vertical and horizontal directions) is used.

As an example, FIG. 13 shows an example in which the output from each of the five vertical and five horizontal cameras when the octopus is the subject is displayed in an easy-to-understand manner. In this example,
There is also parallax in the vertical direction.

In this way, the use of a plurality of cameras whose positions are slightly shifted vertically and horizontally allows the stereoscopic vision to be performed by forming the binocular parallax using the output from one camera as the input to one eye. In addition, when a large number of cameras are used, a person who looks at the display 160 in the output system can obtain a natural stereoscopic view even if the person shakes his / her head.

On the other hand, the three-dimensional volume data indicates the data of each layer of the volume data for three-dimensional stereoscopic image display having a depth layer, and each point in the three-dimensional space has a value. Depending on the appearance of this image data, it can be divided into the following three types.

When not only the surface of the subject but also the image data of the contents are included: The three-dimensional image data obtained by CT scanning or the like has information of the contents for seeing through the contents of the subject and is an image for medical treatment. Often treated as data. When there is image data of only the surface of the subject: When image data of the subject is captured from different angles using multiple cameras and depth is estimated. Therefore, there is no image data of the contents of the subject (this is null data or NULL
Is called). When only the image data of the surface when the subject is viewed from a certain direction is included: The image data of the subject is 1 from a certain angle.
When using one or more cameras to estimate the depth. Therefore, there is only the image data of the surface when the subject is viewed from that direction, and naturally, there is no image data of the contents, and it is null data.

In this way, multi-view stereoscopic image data or 3
The dimensional volume data is expected to be used in various ways as it gives a natural stereoscopic image, and therefore it is necessary to optimize the encoding device.

[0009]

2. Description of the Related Art FIG. 14 shows a conventionally known image data encoding apparatus which operates on the above-mentioned data. The input image data is preferably subjected to motion compensation and preprocessing. It is sent to the orthogonal transformation unit 1 via the pre-processing unit 6 which performs the operation, where DCT, wavelet, Hadamard, KL transformation and the like are performed to transform the image data from the time component signal to the frequency component signal.

That is, in the orthogonal transformation unit 1, when one screen of an input image to be encoded is divided into one-dimensional or multi-dimensional blocks of two or more dimensions and orthogonal transformation is performed,
As shown in FIG. 15, it is divided into low-frequency to high-frequency components. The illustrated example is a two-dimensional block of 8 × 8 pixels.
In the figure, the upper left end is a DC component, and the frequency component is higher toward the lower right.

On the other hand, it is known that image data generally has few high frequency components, and that human eyes hardly perceive the distortion of high frequency components.

Therefore, in a widely known coding apparatus such as standardization, the weighting processing section 2 is provided as a preprocessing section for the quantization or coding section 7 as shown in FIG. The quantized value of the high frequency component is made smaller (coarse) than the quantized value of the low frequency component.

That is, in order to make the quantized value of the high frequency component smaller than the quantized value of the low frequency component, the weighted one called the quantization matrix of ISO standard MPEG2 as shown in FIG. 16 is quantized. To do. This quantization matrix is a pre-process for quantization, and the larger the weight, the larger the calculated quantization value.

Therefore, in the quantizing / encoding unit 7 which sends the encoded result to the receiving side, the quantized value is variable length encoded from the low frequency component toward the high frequency component. Since the information amount is assigned such that the code word becomes smaller as the quantization value becomes smaller, the information amount can be reduced.

Further, as the probability that the frequency component becomes "0" becomes higher, 0 run length (in two-dimensional variable length coding,
There is an effect of the number that "0" continues, that is, the combination of the run length and the code word) and EOB (End Of Block, the value of the frequency component after this code word is not sent in the block), The amount can be reduced.

[0016]

In the conventional image data coding apparatus, the preprocessing method and parameters for quantization such as the quantization matrix are constant in any part of the image within the screen, or the parameters. Was simply switched according to the nature of the image.

However, depending on the application, a high image quality (high resolution) that is clear only in a specified portion of the image is required (it is necessary that the high frequency component is not damaged).
Then, in other parts, it may be possible to use a method of improving the efficiency by lowering the image quality (at the expense of the high frequency component) to facilitate the encoding.

For example, in remote operation manipulation, when performing a work, an operator generally performs the work by aligning the center of the image with the work place. However, in the conventional apparatus, the operator automatically responds to the position of the image. Since the parameters cannot be changed, the coding efficiency is lowered, and the required image quality of the central portion image is deteriorated, which hinders the work.

Therefore, according to the present invention, the frequency component obtained by orthogonally transforming the input image data or the preprocessed input image data by the orthogonal transforming unit is weighted by the weighting processing unit and then quantized or In an image data encoding device that performs encoding processing, an object is to provide a high image quality portion that does not impair high frequency components and a low image quality portion that sacrifices high frequency components in the screen to facilitate encoding as necessary. And

[0020]

In order to achieve the above object, in the image data coding apparatus according to the present invention, as shown in principle in FIG. 1, a horizontal synchronizing signal and a block horizontal synchronizing signal in the apparatus are provided. A block vertical position determining unit 3 for determining the current vertical position of each block forming the input image data, and each block forming the input image data by the vertical synchronization signal and the in-apparatus block vertical synchronization signal. The block horizontal position determining unit 4 for determining the current horizontal position, and the current vertical position and horizontal position of each block by both determining units 3 and 4 and which region in the screen designated by the operator is raised. The parameter calculation unit 5 selects a weighting pattern for the current block according to the parameter indicating whether to make the image quality and gives it to the weighting processing unit 2.

Further, in the above-mentioned present invention, the parameter calculation unit 5 may store a plurality of types of weighting patterns in advance, select any one of the weighting patterns, and give it to the weighting processing unit 2.

Further, in the above-mentioned present invention, a quantization matrix can be used as the weighting pattern.

Further, in the present invention, the parameter calculation unit 5
However, the value for the high frequency component of the quantization matrix can be increased as the distance from the center point of the high image quality region increases.

Further, according to the present invention, the parameter calculation unit 5 and the weighting processing unit 2 forcibly set the higher-order component lower in the frequency components to 0 as the distance from the center point of the high image quality region increases. Is possible.

Further, according to the present invention, the parameter may include the position of the pixel at the center of the area for maintaining the high image quality and the distance from the center.

[0026]

In FIG. 1, the orthogonal transformation unit 1 and the weighting processing unit 2 constitute the known image data coding device shown in FIG. 14, and the present invention improves the weighting processing unit 2 as follows. Is added.

That is, first, the horizontal synchronizing signal, the vertical synchronizing signal,
Further, the current vertical and horizontal positions x and y of each block are detected by the vertical position determining unit 3 and the horizontal position determining unit 4 from a clock signal or the like used in the apparatus by using a counter or the like.

A weighting process as a pre-process of quantization or coding is performed by using the detection results of the decision units 3 and 4 and a parameter given by the operator indicating which position on the screen the image quality is to be enhanced. A weighting pattern (for example, a quantization matrix) in the unit 2 is selected, and a weighting process is performed on the frequency component after the orthogonal transformation in the orthogonal transformation unit 1. Alternatively, with respect to the input image data, the value of the low frequency component is increased to “0” as it goes to the peripheral portion except for the high image quality area.

Therefore, the high frequency component does not deteriorate in the high image quality region, and the high frequency component deteriorates in the other regions as the distance from the center of the high image quality region increases, so that the image quality in the high image quality region can be maintained. , The coding efficiency can be improved.

The above-mentioned image data is a two-dimensional image (monocular to
Basically, the same can be applied to the case of multiview) and the case of three-dimensional volume data. For example,
Even if the stereoscopic vision is not performed by binocular parallax in the multi-lens type and the information is provided in the depth direction by using the three-dimensional volume data, the operator can see the spatial position (device Depending on the situation, the depth is about 1 meter), but it is only an extension from 2D to 3D.

[0031]

Embodiment (No. 1): FIGS. 2 to 4 This embodiment shows an embodiment of a parameter calculation unit 5 used in the image data encoding apparatus according to the present invention shown in FIG. First, an embodiment in which several kinds of quantization matrices prepared in advance are selected according to the position of a block will be described with reference to the flowchart of FIG. In addition, although this embodiment shows the case of a two-dimensional image of a single eye to a multi-eye, it can also be applied to three-dimensional volume data by expanding the number of dimensions.

The size of one screen is A × B as shown in FIG.
As a pixel, the size of a block as a processing unit of image data in one screen is 8 × 8 pixels as usual.

As a parameter given by the operator, the center point of the area where high image quality (resolution) is desired to be maintained is (a, b) as shown in FIG. A parameter (radius) indicating up to which area the high image quality is maintained is r, and a quantization matrix (see FIG. 16) as a weighting pattern to be output is an 8 × 8 array Q _MAT (8,8). ) (Step S
1).

Further, the current block position (pixels on the left shoulder of the block) output by a counter (not shown) used as an embodiment of the vertical position determining unit 3 and the horizontal position determining unit 4 shown in FIG. The position is defined as (x, y) (at step S1).

Then, the rough distance between the current block position (x, y) and the high image quality center point (a, b) is checked,
“U” = | x−a | + | y−b | (S2). However, the distance is originally an average of squared errors, but is set to the sum of absolute values for simplification.

Then, the distance U thus obtained is compared with the parameter r from the operator. If U <r, the quantization matrices Q _MAT (8,8) are all 16 in order to maintain high image quality. / 16 = “1” (at step S3).

As a result, the radius r from the center point (a, b)
When the current block in the image data subjected to the orthogonal transformation is located in the high image quality region, the information of the high frequency component of the block is output without deterioration.

As shown in the area E of FIG. 4, when the distance U is greater than the radius r and is in the area E smaller than (A + B-4 × r) / 6 (step S4), the quantization matrix shown in step S5. Q _MAT (8,8) is used to weight the orthogonal transform coefficient as the frequency component output from the orthogonal transform unit 1 (see FIG. 1).

In this quantization matrix, the harmonic component is heavier in weight, so that a large quantization value is not calculated, rough quantization is performed, and the image is in a low image quality state.

Here, the expression (A + B in step S4
-4 × r) / 6 means that U is from “0” to maximum (A +
B) / 2 is taken and the radius r to (A + B) / 2 (= (A + B-2r) / 2) is equally divided into three and used as a threshold value.

Further, the distance U is (A + B-4 × r) / 6 or more and (A + B−r) /, as shown in the region F of FIG.
If it is smaller than 3 (step S6), the quantization matrix Q _MAT (8,8) shown in step S7 is used. This quantization matrix has a higher weight on the higher-order coefficient side than the quantization matrix in step S5, so that coarse quantization is performed.

Further, as shown in the area G in FIG. 4, the distance U
Is larger than (A + B-r) / 3 (step S
6), the quantization matrix Q _MAT (8, 8
Use 8). This quantization matrix is step S5.
Further, the weight is further increased compared to the quantization matrix of S7 and S7, so that coarse quantization is performed.

As described above, in the above-described embodiments, in these quantization matrices, the closer the position of the block in the screen is to the peripheral portion of the screen, the larger the value of the array of the quantization matrix with respect to the high-frequency component becomes, and the coarser the matrix is. Quantization is performed to obtain a low quality image.

The quantization matrix Q _MAT (8,8) may differ depending on whether the image signal is a luminance signal or a color difference signal, or whether or not inter-frame coding is performed.

FIG. 5 is a block diagram showing a hardware circuit for executing the calculation algorithm shown in FIG.
After outputting the current block position (x, y), the difference between the current block position (x, y) and the parameters a and b designated by the operator is obtained, and the absolute value thereof is taken by the arithmetic unit 11 to be added to obtain U.

On the other hand, (A + B-4 × r) / 6 and (A + B-r) / 3 are calculated from the screen sizes A and B and the parameter r which are given in advance, and these are compared with U and r. By giving it to the device 13, steps S2, S4, S6
The comparison operation corresponding to is performed. The comparison result is passed through a selector 14 composed of a ROM, a switch, etc., and a quantization matrix Q _MAT (8,8) corresponding to the quantization matrix of steps S3, S5, S7 and S8 is stored in memories 15a to 1a.
Select from 5d and output.

Embodiment (No. 2): FIGS. 6 to 7 FIG. 6 shows a modification of the embodiment (No. 1), which also shows an embodiment of the parameter calculation unit 5, and the quantization matrix Q. An example of calculating _MAT (8,8) according to the position of the screen is described.

First, as in the first embodiment, the size of one screen is set to A × B pixels as shown in FIG. 3, and the size of a block as a processing unit of image data in one screen is usually set. As shown in FIG.

Further, as a parameter given by the operator, the center point of the area where the high image quality (resolution) is desired to be maintained is (a, b), and the area from the center point (a, b) maintaining the high image quality Let r be the parameter (radius) that indicates whether to keep the image quality as high, and the quantization matrix to be selected is an 8 × 8 array Q _MAT (8,8), and further away from the center point (a, b). The parameter indicating the degree
(Step S11).

The current block position (the position of the pixel on the left shoulder of the block) is set to (x, y) (at step S11).

Then, the rough distance between the current block position (x, y) and the high image quality center point (a, b) is checked,
“U” = | x−a | + | y−b |, and k = 2 ×
(U-r) / (A + B-2r) is determined (step S1)
2).

The meaning of the equation of k is that, if U <r, the high image quality is maintained, so if it is not taken into consideration, the maximum value that U can take is approximately (A +
B) / 2, the range of fluctuation of U is r <U <(A + B) / 2. However, when the fluctuation range of U is normalized from “0” to “1”, 0 < U-r <{(A + B) / 2} -r.

When this is modified, 0 <2 × (U−r) / (A + B−2r) <1. Therefore, if 2 × (U−r) / (A + B−2r) is replaced with k, this k Indicates how far the current block is from the center. In this case, 0 <k
<1.

Therefore, it is judged whether or not k <0 (step S13), and when k <0, it means that U <r. Therefore, a limiter is applied so that k = 0 ( Step S14). Therefore, all the values of the quantization matrix in step S15 become "1", and the data after orthogonal transformation is output as it is.

On the other hand, if k ≧ 0, the quantization matrix Q
_MAT (8,8) is determined as shown in step S15. For each value of this quantization matrix, the larger k is,
The higher the frequency component, the larger the value.
The farther the current block is from the center, the larger the value of the array of the quantization matrix for the high frequency component, so that the data after the orthogonal transformation is coarsely quantized.

FIG. 7 is a block diagram showing a hardware circuit for executing the calculation algorithm shown in FIG.
First, U is calculated by using the current positions x and y of the block and the parameters a and b from the operator and passing through the absolute value calculation units 11 and 12.

Then, k is calculated from the division unit 16 using this U, the parameter r from the operator, and the sizes A and B of one screen. The value k is fixed to 0 when it becomes 0 or less by the limiter 17.

After that, based on k from the limiter 17,
1+ constituting Q _MAT (8,8) shown in step S15
k, (1 + k) ² , (1 + k) ³ , 1 + k / 2, 1 + k
Five kinds of values of / 4 are calculated and given to the switch 18.

These calculation results are obtained by first giving the position of each pixel in the block currently processed by a counter or the like to the memory 15 so that the memory 15 can quantize the matrix Q _MAT (8,8) according to the pixel position. ) Switch 18 because it remembers which value to select
Can be switched. Since this quantization matrix is normalized, it is not necessary to divide "16" with respect to the output from the switch 18, and the output may be output as it is.

Embodiment (3): FIGS. 8 to 9 FIG. 8 shows a further modified example.
A method in which the parameter calculation unit 5 and the weighting processing unit 2 are mixed and combined, and a method of setting the high frequency component (higher order coefficient) obtained by the orthogonal transformation according to the position of the image block to “0” is adopted. There is.

First, as in the first and second embodiments, the size of one screen is set to A × B pixels as shown in FIG. 3 and the size of a block as a processing unit of image data in one screen. 8 × 8 pixels as usual.

As a parameter given by the operator, the central point of the area where the high image quality (resolution) is desired to be maintained is (a, b), and the area from the central point (a, b) where the high image quality is maintained to Let r be a parameter (radius) indicating whether or not to keep the high image quality (step S21).

Further, assuming that the current block position (the position of the pixel on the left shoulder of the block) is (x, y), an image block after orthogonal transformation which is an input to the weighting processing unit 2 shown in FIG. 1, that is, an input image Let the block be C (8,8) and the image block output from the weighting processing unit 2 be D (8,8) (at step S21).

Further, the scan order when the two-dimensional DCT is one-dimensionally expanded (normally, scanning is performed from the low frequency component to the high frequency component, but in the present embodiment, the scan order is zigzag scan) is shown in the figure. As J
It is given as an array of (8, 8).

Then, a threshold value T for making the following determination
H is set to TH1 to TH4 (at step S21).

After setting in this way, the rough distance between the current block position (x, y) and the high image quality center point (a, b) is checked, and "U" = | x-a | + | y- b | and compared with the above parameter r (at step S22), and if U <r, the threshold value TH = TH1 is set (at step S23).

If U ≧ r and smaller than (A + B-4 × r) / 6 as shown in the area E of FIG.
4) and the threshold value TH is set to TH2 (at step S25). This (A
The meaning of the expression + B-4 × r) / 6 is as shown in FIG.
U takes a value of 0 to (A + B) / 2 and r to (A + B) / 2.
B) / 2 is a threshold value when it is divided into three equal parts (= (A + B-2r) / 2).

If U is (A + B-4 × r) / 6 or more and smaller than (A + B−r) / 3 in step S24 (S26), the threshold value TH = It is set to TH3 (at step S27).

Otherwise (region G in FIG. 4), the threshold value TH = TH4 is set (at step S28).

Next, the above array of J (8,8) (i,
i, j in j) is initialized to i = 1, j = 1 (steps S29, S30), and this J (i, j) is compared with the threshold TH (step S31).

As a result, the value of each element of the input image block C (8, 8) at the position where the value of the element by J (8, 8), that is, the value of the scan order exceeds the threshold value TH is set to "0". (S33), when the threshold value TH is not exceeded, the value of the input image block C (8,8) is directly output image block D.
It is given as (8, 8) to the quantization / encoding unit of the next stage (see FIG. 14) (at step S32). Then, i and j are incremented and executed until i = 8 and j = 8 (S34 to S37).

Therefore, in this embodiment, the threshold value TH1
When>TH2>TH3> TH4, the threshold TH becomes smaller as it goes to the peripheral portion of the screen (from region E to G in FIG. 4). Therefore, the value of the scan order of J (8,8) is the threshold value. Since it often becomes larger than TH, the component which removes the value of each pixel of the input image block C (8, 8) subjected to the orthogonal transformation by “0” is closer to the low frequency component.

FIG. 9 is a block diagram showing a hardware circuit for executing the calculation algorithm shown in FIG.
First, U is calculated by using the current positions x and y of the block and the parameters a and b from the operator and passing through the absolute value calculation units 11 and 12.

Then, U and r, (A + B-4 × r) /
6 or (A + B−r) / 3 respectively in the comparator 13. The result of this comparison is input to the selector 14 ', and one of the threshold values TH1 to TH4 is selected and output.

The output result of the selector 14 'is the scan array (i, j) of J (8,8) output from the memory 22 by the counters 20 and 21 based on the clock generated for each pixel in the device. The elements are input to the comparator 19 together with the elements, the comparison result is input to the selector 23, and the selector 23
Either "0" or the element of the input image block C (8,8) is selected and quantized as an image block D (8,8) /
Output to the encoding unit.

Embodiment (No. 4): FIGS. 10 to 11 FIG . 10 shows still another modification. This embodiment includes the above-mentioned embodiment (No. 2) and embodiment (No. 3). A combination of the parameter calculation unit 5 and the weighting processing unit 2
And are mixed.

First, similarly to each of the above-described embodiments, the size of one screen is set to A × B pixels as shown in FIG. 3, and the size of a block as a processing unit of image data in one screen is set to a normal size. As described above, the number of pixels is set to 8 × 8 (step S41).

As a parameter given by the operator, the center point of the area where high image quality (resolution) is desired to be maintained is (a, b), and the area from the center point (a, b) where high image quality is maintained Let r be a parameter (radius) that indicates whether to keep the high image quality (S41).

Furthermore, the current block position (the position of the pixel on the left shoulder of the block) is (x, y), and the center point (a, b)
Assuming that the parameter indicating the degree of separation from k is k
The image block after the orthogonal transformation, which is the input to the weighting processing unit 2 shown in FIG. 2, that is, the input image block is C (8, 8),
The image block output from the weighting processing unit 2 is D
(8, 8), and the scan order (zigzag scan order) when one-dimensionally expanding the two-dimensional DCT is given as an array of J (8, 8) as shown in the drawing (at step S41).

Then, the rough distance between the current block position (x, y) and the high image quality center point (a, b) is checked,
“U” = | x−a | + | y−b |, and k = 2 × (U−r) / (A + B−2r) as in the second embodiment.
(S42).

Then, when k <0 is satisfied (the same S4
3), since U <r, a limiter is applied to k = 0 (at step S44).

The process up to this point is the same as that of the embodiment (part 2). In this embodiment, the process of the embodiment (part 3) is performed in step S4.
5 to S53 are executed. These steps S45-S
Reference numeral 53 corresponds to steps S29 to S37 of the embodiment (part 3).

However, in the present embodiment, in step S47 instead of step S31 of the embodiment (part 3), k
And α (0 <α <64) are multiplied to form an integer (INT),
The value is subtracted from the maximum value "64" of the scan order, and J
The value (scan order value) of the element (8, 8) is compared with the above calculation result.

When the value of the element of J (8,8) is large, the output of the processing unit 2 is "0", otherwise, the value of the frequency component of the input image block C (8,8) is output as it is. (S48, S49 of the same).

In this way, in this embodiment, the smaller the k and the smaller α, the higher the frequency component is not reduced to zero.

FIG. 11 is a block diagram showing a hardware circuit for executing the calculation algorithm shown in FIG. 10, in which the positions x and y of the blocks and the parameters a and b from the operator are used. U is calculated, and k is calculated by the division unit 16 using U, the parameter r from the operator, and the image sizes A and B. The value k is set to 0 when it becomes 0 or less by the limiter 17.

After that, k is multiplied by α, and the result is subtracted from “64”. This and the counters 20 and 2
The values of the elements of J (8,8) according to 1 and 22 are compared by the comparator 19, and whether the selector 23 is the element of the input image block C (8,8) to the processing unit 2 based on the comparison result. 0 is selected as the output image block D (8,8).

[0088]

As described above, according to the image data encoding apparatus of the present invention, the current vertical position and horizontal position of each block constituting the input image data and the screen instructed by the operator. Of the high-frequency component with respect to the input image data as much as possible by giving a weighting pattern for the current block to the weighting processing unit according to the parameter indicating which region of the high-quality image. Since the parameter calculation unit for "." Is provided, it is possible to maintain the image quality of the central image which is important for the operator's work and to reduce the image quality of the peripheral image necessary to know the general situation. Efficiency can be increased.

Further, by improving the coding efficiency, the image quality can be improved as compared with the conventional case. At present, in encoding, which is the mainstream of standardization of high-efficiency image encoding (CCITT, ISO, etc.), it can be expected that efficiency will be considerably improved due to such deterioration in image quality.

[Brief description of drawings]

FIG. 1 is a block diagram showing the principle of an image data encoding device according to the present invention.

FIG. 2 is a flowchart showing an embodiment (part 1) by software of the image data encoding device according to the present invention.

FIG. 3 is an explanatory view (1) of the principle of the embodiment (No. 1).

FIG. 4 is an explanatory view (2) of the principle of the embodiment (No. 2).

FIG. 5 is a block diagram of a hardware circuit of the embodiment (No. 1).

FIG. 6 is a flowchart showing an embodiment (part 2) by software of the image data encoding device according to the present invention.

FIG. 7 is a block diagram of a hardware circuit of an embodiment (No. 2).

FIG. 8 is a flowchart showing an embodiment (part 3) by software of the image data encoding device according to the present invention.

FIG. 9 is a block diagram of a hardware circuit according to an embodiment (No. 3).

FIG. 10 is a flowchart showing an embodiment (part 4) by software of the image data encoding device according to the present invention.

FIG. 11 is a block diagram of a hardware circuit of an embodiment (No. 4).

FIG. 12 is a block diagram showing a general configuration of a multi-view stereoscopic image system.

[Fig. 13] Fig. 13 is a diagram showing an example of multi-lens (5 eyes x 5 eyes) camera output.

FIG. 14 is a block diagram showing a conventional example.

FIG. 15 is a diagram for explaining coefficient components of image data after orthogonal transformation.

FIG. 16 is a diagram showing an example of a conventionally known quantization matrix.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Orthogonal transformation part 2 Weighting processing part 3 Vertical direction position determination part 4 Horizontal direction position determination part 5 Parameter calculation part 6 Pre-processing part 7 Quantization / coding part In the figure, the same code | symbol shows the same or equivalent part.

Claims (6)

[Claims] 1. A weighting processing unit (2) performs a weighting process on a frequency component obtained by orthogonally transforming input image data or preprocessed input image data by an orthogonal transformation unit (1). In an image data encoding device that performs quantization or encoding processing, a block vertical position determination that determines the current vertical position of each block that constitutes the input image data by a horizontal synchronization signal and a block horizontal synchronization signal in the device A block lateral position determining unit (4) that determines the current horizontal position of each block that composes the input image data based on the vertical synchronizing signal and the in-device block vertical synchronizing signal; Weighting for the current block by the current vertical position and horizontal position of each block by (3, 4) and the parameter that indicates which area in the screen the high image quality is specified by the operator. Image data encoding apparatus characterized by comprising parameter calculation unit which selects the pattern giving to the weighting processing unit (2) and (5), a. 2. The parameter calculation unit (5) stores a plurality of types of weighting patterns in advance, and selects one of the weighting patterns to give to the weighting processing unit (2). Item 1. The image data encoding device according to Item 1. 3. The image data encoding device according to claim 2, wherein the weighting pattern is a quantization matrix. 4. The image data according to claim 4, wherein the parameter calculation unit (5) increases the value for the high frequency component of the quantization matrix as the distance from the center point of the high image quality region increases. Encoding device. 5. The parameter calculation unit (5) and the weighting processing unit (2) forcibly set lower higher-order components of the frequency components to 0 as the distance from the center point of the high image quality region increases. The image data encoding device according to claim 1, wherein 6. The method according to claim 1, wherein the parameter includes the position of a pixel at the center of the area where the high image quality is maintained and the distance from the center. Image data encoding device.