JP2001112007A - Image compressor and image expander - Google Patents

Abstract

PROBLEM TO BE SOLVED: To enhance the compression rate in no loss compression of a three- dimensional image signal. SOLUTION: A three-dimensional optimum adaptive encoding means 100 receives a three-dimensional digital image signal, and a pixel correlation evaluation means 101 calculates the evaluation of the correlation among prescribed respective pixels that can be referred to belonging to current referenced pixels and elapsed referenced pixels of a prediction object pixel. An optimum value calculation means 102 obtains a set of selected pixels or a proper optimum function maximizing the correlation between the pixels on the basis of the evaluation of the correlation to calculate an optimum value. An optimum adaptive encoding means 103 encodes the object pixel on the basis of the calculate optimum value to generate an encoded signal. An entropy encoding means 200 applies entropy encoding to the encoded signal to generate a three- dimensional no loss compression image signal.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image compression device and an image decompression device, and more particularly to a plurality of screens having a temporal correlation, such as a still image continuous image and a moving image.
The present invention relates to an image compression device that performs lossless image compression of a two-dimensional image and an image expansion device that expands the compressed image signal.

[0002]

2. Description of the Related Art Recently, the field of multimedia has been developed, and a large amount of images such as still images and moving images are flowing through networks. In addition, high-resolution digital still cameras have become widespread, and the image content market has become enormous. In archiving, distributing, editing, and the like of these images, there is a demand for lossless compression of image quality, particularly in broadcast stations and medical fields, without deterioration. In particular, the amount of information of a three-dimensional image composed of a plurality of screens having a temporal correlation such as an image continuously shot by a still image camera and a general moving image (an image shot by a video camera, an animation, etc.). It is indispensable to increase the compression ratio because of the huge volume. Various methods are known as a lossless image compression method that can retain the original information even after such a compression / expansion process. One of these methods is a three-dimensional predictive DPCM process. Will be described.

FIG. 18 is a configuration diagram of a conventional image compression apparatus that performs a three-dimensional prediction type DPCM process. A conventional image compression apparatus includes a three-dimensional prediction type DPCM 50 for calculating a prediction error.
0 and entropy encoding means 200 for performing entropy encoding of the prediction error. 3D prediction type D
The PCM 500 receives a three-dimensional digital image signal, and calculates a prediction error which is a difference signal between a target pixel and a prediction value adaptively calculated within a screen and between screens. Also, the entropy encoding means 200 is a three-dimensional prediction type DPCM5
The prediction error calculated in 00 is subjected to variable-length coding by Huffman coding or arithmetic coding, and a compressed code is output.

DPC performed by three-dimensional prediction type DPCM500
The M encoding will be described. In the DPCM encoding, a difference signal d_i is calculated from a target pixel x_i and a prediction value p_i predicted from its surrounding pixels or past pixels. A specific example will be described. FIG. 19 shows a pixel correspondence in a three-dimensional image. For example, in a region that can be regarded as a still image in motion detection, the prediction value p of the target pixel P
Is calculated using the temporal reference pixel x′0.

[0005]

## EQU00001 ## p = x'0 (1) For example, in a region regarded as a moving image, prediction processing is performed using pixels around P. In the case of the prediction process in which the adaptive process is performed, the pixel value with the smallest error from the target pixel, the value obtained by calculating the average value of some pixels,
For example, a predicted value is selected. At this time, a special code indicating the selected calculation mode is added. For example, the special code SV when p = x1 is selected is SV = 1, and p = y
When 1 is selected, SV = 2, and so on. This special code is generally used to identify some suitable scan block,
For example, it is added as header information for each scan line.

[0006] As a predicted DPCM output, a difference signal d_i is obtained by the following equation.

[0007]

D_i = x_ip−p_i (2) Here, the predicted value p_i is a prediction function selected by the adaptive processing described above.

[0008]

However, the conventional image compression apparatus for performing three-dimensional lossless compression has a problem that the compression ratio is low and is not suitable for practical use.

In the three-dimensional lossless compression described above, the calculation mode is made variable for each appropriate scan block, for example, for each scan line. Since such processing is not optimal processing according to the characteristics of an actual image, there are cases where prediction accuracy is poor. In the same scan block, a fixed calculation mode is set, and a difference signal is increased depending on a pixel, and a case where prediction accuracy is not good occurs.
In the adaptive processing in units of scan blocks, the contribution to the improvement of the compression ratio is reduced. As described above, in the conventional lossless compression, the prediction accuracy cannot be improved, and therefore, it has been difficult to improve the compression ratio.

On the other hand, when adaptive processing is performed for each pixel and a special code is added to each pixel, the number of information bits increases, so that the compression ratio is not necessarily improved. As described above, the conventional lossless compression technique has hardly been widely used because of its low compression ratio.

The present invention has been made in view of the above points, and an image compression apparatus capable of increasing a compression ratio in three-dimensional lossless compression, and a compressed image signal compressed by the image compression apparatus. An object of the present invention is to provide an image decompression device for decompression.

[0012]

According to the present invention, in order to solve the above-mentioned problem, a three-dimensional lossless image composed of a plurality of screens having a temporal correlation such as a continuous image of a still image and a moving image is provided. In the image compression device for performing compression, a three-dimensional digital image signal is input, and pixel correlation evaluation means for evaluating a correlation of each of predetermined referenceable pixels belonging to a current reference pixel and a temporal reference pixel of a prediction target pixel, and Optimal value calculating means for calculating an optimum value for maximizing the correlation between pixels based on pixel correlation evaluation, and optimal adaptive coding means for coding a target pixel based on the optimum value and generating a coded signal And an entropy encoding unit for entropy encoding the encoded signal.

The image compressing apparatus having such a configuration receives a three-dimensional digital image signal, and the pixel correlation evaluation means correlates the predetermined referenceable pixels belonging to the current reference pixel and the temporal reference pixel of the prediction target pixel. Is calculated, and the optimum value is calculated by, for example, obtaining a selected set of pixels or an appropriate optimum function to maximize the correlation between pixels by the optimum value calculating means. The optimal adaptive encoding means encodes the target pixel based on the calculated optimal value. The entropy coding means performs entropy coding on the coded signal to generate a three-dimensional lossless compressed image signal.

In an image decompression apparatus for decompressing a compressed image signal obtained by losslessly compressing a three-dimensional image composed of a plurality of screens having a temporal correlation such as a continuous image of a still image and a moving image, the optimum adaptation is provided. Entropy decoding means for decoding the image compression signal which has been entropy-encoded after encoding, and pixel correlation evaluation for evaluating the correlation between individual referenceable pixels belonging to the current reference pixel and the temporal reference pixel of the target pixel Means, an optimum value calculating means for calculating an optimum value for maximizing a correlation between pixels based on the pixel correlation evaluation, and an optimum adaptation for decoding a target pixel from a coded signal based on the optimum value. An image decompression device comprising: a decoding unit.

An image decompression device having such a configuration receives a three-dimensional compressed image signal compressed by the image compression device and performs entropy decoding of the compressed image signal by entropy decoding means. The pixel correlation evaluation unit evaluates the correlation of each of the predetermined referenceable pixels belonging to the current reference pixel and the temporal reference pixel of the target pixel, and the optimum value calculation unit selects a pixel that maximizes the correlation between the pixels. Alternatively, an optimal value is calculated by finding an appropriate optimal function. The optimal adaptive decoding means decodes the compressed image signal subjected to the entropy decoding based on the optimal value, and decodes the target pixel.

[0016]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a configuration diagram of an image compression device according to an embodiment of the present invention.

An image compression apparatus according to the present invention inputs a three-dimensional digital image signal composed of a plurality of screens having a temporal correlation and performs an optimal adaptive process to encode the three-dimensional digital image signal. Three-dimensional optimal adaptive coding means 10
0 and entropy coding means 200 for further entropy coding the coded image signal.

The three-dimensional optimal adaptive coding means 100 includes a pixel correlation evaluation means 101 for evaluating the correlation of each of the predetermined referenceable pixels of the prediction target pixel, and a pixel correlation evaluation means 10.
The optimum value calculating means 102 calculates an optimum value based on the pixel correlation evaluation of No. 1 and the optimum adaptive coding means 103 codes a target pixel based on the optimum value.

The pixel correlation evaluation means 101 evaluates the correlation between each of the predetermined referenceable pixels belonging to the current reference pixel and the temporal reference pixel of the prediction target pixel. The three-dimensional image will be described. FIG. 2 shows a pixel correspondence of a three-dimensional image in the image compression apparatus according to the embodiment of the present invention. As shown in FIG.
The two-dimensional image can be regarded as a predetermined screen arranged in time series. For example, assuming that the image to which the pixel to be predicted belongs is the current reference image, before and after that, there are an image that becomes temporally past and an image that becomes future. In this manner, the reference image temporally separated from the current reference image is set as the temporal reference image. For example, in a region that can be regarded as a still image by motion detection, the correspondence between pixels of the current reference image and images of the temporal reference image is indicated by a dotted line. Pixels x1, x2, and y0 of the current reference image shown in the figure correspond to x'1, x'2, and y'0 at the same position of the temporal reference image 1. In this way, predetermined pixels separated by time can be set as reference pixels. Also, a predetermined pixel around the prediction target pixel in the current reference image can be set as a reference pixel. For example, assuming that x1 of the current reference image shown in the figure is the prediction target pixel, x'1 of the temporal reference pixel and x2, y1, etc., which are pixels around x1 of the current reference image can be set as reference pixels. As described above, the pixel correlation evaluation unit 101 evaluates the correlation of each of the predetermined referenceable pixels belonging to the current reference pixel and the temporal reference pixel of the prediction target pixel. The predetermined reference pixel is a pixel that satisfies a predetermined condition among the reference pixels described above, for example, a pixel that is in a predetermined geometric structure area determined by a directional feature of the current reference image.

The correlation evaluation uses similarity or distance, which is a general correlation calculation. For example, a set-theoretic similarity is selected as an example of the similarity. Assuming that the current reference pixels that can be referred to and selected by a predetermined condition are xn, yn, zn, and others, and that the temporal reference pixel is t, m sets of similarities Sm are:

[0021]

Sm = (min ｛xn, yn, zn, t ,,｝) / (max ｛xn, yn, zn, t ,,｝) <= 1 (3) . The higher the value, the higher the similarity,
At 1 the correlation is maximum. n is an appropriate selected number, and the surrounding pixels allow multiple selections. If Chebyshev distance is selected as an example of the distance, m sets of distances Dm are:

[0022]

Dm = Σ | Δm ｛xn, yn, zn, t,...｝ |> = 0 (4) Here, Δm ｛｝ is a function for obtaining m sets of difference values between pixels, and the smaller the value, the higher the correlation, and the correlation becomes maximum at 0.

Returning to FIG. Optimal value calculation means 1
In step 02, an optimum value is calculated by obtaining a selected set or an appropriate optimum function that maximizes the correlation between pixels calculated by the pixel correlation evaluation means 101. In the case of the similarity represented by Expression (3), the maximum correlation is max {Sm}, and in the case of the distance represented by Expression (4), the maximum correlation is min {Dm}. Hereinafter, the similarity will be described. In the case of the distance, it is an inverse function of the similarity. In the case of a selection type that selects a pixel set that maximizes the correlation, the optimal value SP is

[0024]

SP = Select {xn, yn, zn, t,..., {For max} Sm} (5) Here, “Select” indicates selection of a pixel set. In addition, the optimal value SP ′ for the arithmetic type is

[0025]

SP ′ = round [f (Select {xn, yn, zn, t,..., {For max} Sm})] (6) round {f (x)} represents conversion of a certain function f (x) into an integer. f () is an optimal function for calculating an optimal value under a condition of for or less. Also, SP and SP 'can be restricted by the attribute a. The optimal value limited by the attribute a is SP '' _ a
Is

[0026]

SP ″ _a = SP_a or SP′_a (7) The attribute a includes a distance weight to the peripheral pixel value, a geometric weight of the peripheral pixel, and the like. here,
It is assumed that the processing is to apply an appropriate weight restriction to some peripheral pixels.

The optimum adaptive coding means 103 performs coding processing based on the optimum value calculated by the optimum value calculation means 102. The encoding process includes a predictive DPCM method, a one-layer subband conversion method, a multi-layer one-layer subband conversion method, and the like. The details of each will be described later.

The entropy coding means 200 performs an entropy coding process on the coded signal output from the three-dimensional optimal adaptive coding means 100 to generate a three-dimensional lossless compressed image signal. Huffman codes, Golomb-Rice codes, arithmetic codes, and the like are known as entropy coding methods. The image compression apparatus according to the present invention selects an appropriate one of these methods.

The operation of the image compression apparatus having such a configuration will be described. When a three-dimensional digital image signal is input to the three-dimensional optimal adaptive encoding means 100, the pixel correlation evaluation means 1
01, the correlation evaluation of each predetermined reference pixel is performed.
First, motion detection is performed around the prediction target pixel. In the motion detection, for example, a difference between the current reference image and a pixel of the temporal reference image in the t direction separated by time is obtained, and the evaluation is performed.
As a result of the motion detection, if the area is almost a still image,
A pixel in the t direction is added as a reference pixel of the prediction target pixel. On the other hand, in a region that can be regarded as a moving image, only predetermined peripheral pixels in the same image are set as reference pixels. Also, after local motion position correction by motion vector calculation, motion detection can be performed again. This is an optimal process for a moving local part of the temporal reference image. The optimum value calculation means 102 receives the calculated correlation evaluation, finds a selected set or an appropriate optimum function that maximizes the correlation between pixels, calculates an optimum value, and outputs the optimum value to the optimum adaptive coding means 103. . The optimal adaptive encoding unit 103 encodes a three-dimensional digital image signal based on the optimal value. The entropy of the signal encoded by the optimal adaptation process described above is reduced. The coded signal having the reduced entropy is converted into an entropy coded unit 200
, A lossless compressed image signal having a high compression ratio is generated.

Next, an image decompression device for decompressing a three-dimensional compressed image signal compressed by the above-described image compression device will be described. FIG. 3 is a configuration diagram of an image decompression device according to an embodiment of the present invention.

An image decompression device according to the present invention comprises an entropy decoding means 400 for entropy decoding a three-dimensional compressed image signal compressed by the above-described image compression device.
And a three-dimensional optimal adaptive decoding means 300 for decoding the decoded signal by optimal adaptive processing.

The three-dimensional optimal adaptive decoding means 300 evaluates the correlation of each of the predetermined referenceable pixels of the target pixel, and calculates the optimum value based on the pixel correlation evaluation of the pixel correlation evaluation means 301. It comprises an optimum value calculating means 302 for calculating, and an optimum adaptive decoding means 303 for decoding the target pixel based on the optimum value.

The pixel correlation evaluation means 301 calculates a pixel correlation evaluation for each of predetermined referenceable elements. The evaluation of the pixel correlation is performed by the pixel correlation evaluation means 10 of the image compression apparatus described above.
1 is the same as that of the first embodiment, and the same evaluation result as that of the image compression apparatus is obtained. In general, since reference pixels existing near the frame of an image are restricted, processing that does not perform optimal processing is substituted. The optimum value calculating means 302 selects a pair having the maximum correlation based on the evaluation of the pixel correlation, and calculates an optimum value.
The calculation of the optimum value is the same processing as that of the above-described optimum value calculating means 102 of the image compression apparatus, and obtains the same optimum value as that of the image compression apparatus. As described above, the optimal adaptive arithmetic processing performed by the pixel correlation evaluation unit 301 and the optimum value calculation unit 302 of the image decompression device according to the present invention is performed by the pixel correlation evaluation unit 101 and the optimum value calculation unit 102 of the image compression device according to the present invention. Perform the same processing as the optimal adaptive calculation processing performed by.

The optimum adaptive decoding means 303 outputs the output signal of the entropy decoding means 400 and the optimum value calculating means 30
The target pixel is decoded from the optimum value calculated in step 2. The entropy decoding unit 400 performs entropy decoding of the three-dimensional compressed image signal compressed by the above-described image compression device, and outputs this to the three-dimensional optimal adaptive decoding unit 300.

The operation of the image decompression device having such a configuration will be described. The three-dimensional compressed image signal is input, entropy-decoded by the entropy decoding means 400, and output to the three-dimensional optimal adaptive decoding means 300. The pixel correlation evaluation unit 301 evaluates the correlation of each of the predetermined referenceable pixels of the prediction target pixel, and the optimum value calculation unit 302 calculates the optimum value. This optimal adaptation process
The processing is the same as the processing performed by the image compression apparatus described above, and the same result is obtained. The optimal adaptive decoding means 303 decodes the target pixel from the obtained optimal value and the three-dimensional compressed image signal entropy-decoded by the entropy decoding means 400.

The optimum adaptive processing and encoding means will be described with a specific example. Hereinafter, if the optimal adaptive encoding unit 103 is a DPCM encoding unit that performs DPCM encoding first, if it is a single-layer subband conversion unit that performs subband encoding, and thirdly, the image signal is A case of multi-layer one-layer sub-band conversion in which sub-band coding is performed by dividing into layers will be described. In addition, the pixel correlation evaluation unit 301 and the optimum value calculation unit 3 of the image decompression device
02 is the same as the pixel correlation evaluation means 101 and the optimum value calculation means 102 of the image compression apparatus, and will not be described, and the following description will be made mainly for the image compression apparatus. When describing the optimal adaptive encoding process of the image compression device, the optimal adaptive decoding process of the image decompression device will be described.

The first optimal adaptive coding means 103 is a DPC
The case of the M encoding means will be described. here,
Optimal D for selecting any referenceable pixel and calculating the optimal value
The PCM method and the fan-optimal DPCM method for limiting the referenceable pixels to a predetermined geometric structure area will be described.

First, the optimal DPCM method will be described. FIG. 4 shows an optimal DPCM according to an embodiment of the present invention.
1 is a configuration diagram of an image compression device that performs the following. The same components as those in FIG. 1 are denoted by the same reference numerals, and description thereof is omitted. This image compression apparatus has a three-dimensional optimal DPCM encoding unit 110a that inputs a three-dimensional digital image signal and performs DPCM encoding.
The three-dimensional optimal adaptive coding unit 100a includes a pixel correlation evaluation unit 101a that evaluates the correlation of each of the referenceable pixels of the prediction target pixel, and an optimum value that calculates an optimum value based on the pixel correlation evaluation of the pixel correlation evaluation unit 101a. Calculation means 102a
And an optimal DPCM encoding unit 113 for DPCM encoding the target pixel based on the optimal value.

The pixel correlation evaluation means 101a, like the pixel correlation evaluation means 101, evaluates the correlation of each of the predetermined referenceable pixels belonging to the current reference pixel and the temporal reference pixel of the prediction target pixel. Here, the referenceable pixels around the prediction pixel in the current reference image to which the prediction pixel belongs, and further, if no motion is detected around the prediction pixel, the corresponding pixel of the temporal reference image is referred to as the referenceable pixel. And
At the time of decoding, pixels that can be referred to for prediction calculation are pixels for which decoding has been completed (past pixels, previous pixels, and rear pixels).
Only. Therefore, this restriction also applies to encoding. Among the pixels that can be referred to, an arbitrarily selected pixel is subjected to individual correlation evaluation based on the similarity of Expression (3), the distance of Expression (4), and the like. A description will be given of the similarity. An appropriate peripheral pixel of the prediction target pixel x_i is x_i
_J and the temporal reference pixel x_i_t, the similarity Sm
From the equation (3),

[0040]

Sm = (min {x_i_j, x_i_t}) / (max {x_i_j, x_i_t}) (8) Since the distance has an inverse function relationship with the similarity, the description is omitted.

The optimum value calculating means 102a, like the optimum value calculating means 102, obtains a selected set or an appropriate optimum function that maximizes the correlation between pixels based on the evaluation value calculated by the pixel correlation evaluating means 101a. To calculate the optimal value.
In the case of the selection type, the optimum value is expressed as in Expression (5). In the case of the arithmetic type, it is expressed as in equation (6). In many cases, the optimal function f () included in Equation (6) is the weight w_i_j of the determined multiple reference pixels x_i_j and x_i_t.
j, w_i_t can be treated as a linear combination. In this case, the predicted value p_i is

[0042]

(9) p_i = round {[i, j]} w_i_j * x_i_j + [i, t] {w_i_t * x_i_t} (9) Also, weights w_i_j, w_i_
t is normalized,

[0043]

[I, j] Σw_i_j + [i, t] Σw_i_t = 1 (10)

The optimum DPCM coding means 113 calculates a difference signal d_i between the prediction value p_i calculated by the optimum value calculation means 102a and the pixel x_i to be predicted, and performs coding. The encoding is

[0045]

(11) d_i = x_ip-p_i (11) Further, the decoding process for the equation (11) is as follows.

[0046]

X_i = d_i + p_i (12)

The operation of the image compression apparatus having such a configuration will be described. Three-dimensional optimal DPCM encoding means 110a
When a three-dimensional digital image signal is input to the pixel, a pixel correlation evaluation unit 101a performs a correlation evaluation for each pixel that can be referred to. As a result of the motion detection, if it is a region that can be regarded as a still image, pixels of the temporal reference image in the t direction separated from the current reference image by time are added as referenceable pixels. Only the peripheral pixels of are used as reference pixels. Also, after local motion position correction by motion vector calculation, motion detection can be performed again. The optimum value calculation means 102a calculates the optimum value by, for example, Expression (9). The optimal DPCM encoding means 113 is given by the following equation (11).
The DPCM encoding is performed as follows. Since the entropy of the signal subjected to the DPCM encoding is reduced by the optimal adaptive processing, the entropy encoding by the entropy encoding unit 200 generates a lossless compressed image signal having a high compression rate. Further, the image decompression device performs the same optimal adaptation processing as described above, and obtains the DPCM by Expression (12).
Perform decryption.

Next, the fan optimum DPCM method will be described. FIG. 5 is a configuration diagram of an image compression apparatus that performs fan-optimal DPCM according to an embodiment of the present invention. 1 and 4 are denoted by the same reference numerals, and description thereof is omitted. This image compression apparatus has a three-dimensional fan optimal DPCM encoding unit 110b that inputs a three-dimensional digital image signal and performs DPCM encoding.

Three-dimensional fan optimal adaptive coding means 100b
Is a fan pixel correlation evaluation unit 101 that evaluates the correlation of individual referenceable pixels within a predetermined geometric structure area (hereinafter, referred to as a fan area) determined by the directional characteristics of an image.
b, an optimum value calculation unit 102b that calculates an optimum value based on the pixel correlation evaluation of the fan pixel correlation evaluation unit 101b,
Optimum D for subject pixel to DPCM encoding based on optimal value
And PCM encoding means 113.

Similarly to the pixel correlation evaluation means 101, the fan pixel correlation evaluation means 101b evaluates the correlation of each of the predetermined referenceable pixels belonging to the current reference pixel and the temporal reference pixel of the prediction target pixel. Here, it is assumed that an element in the fan area of the current reference image to which the predicted pixel belongs can be referred to. It is assumed that the fan area exists only in the current reference image.

The fan area will be described. FIG. 6 is a conceptual diagram of the fan area. The fan region is a geometric structure region uniquely determined by a plurality of directional features of an image and determined by the direction, and the fan regions do not interfere with each other. That is, any fan area is independent. A case where this is approximately established is called an approximate fan, and in an ordinary application, an approximate fan is handled.
In the following description, the fan area indicates an area including an approximate fan.

An appropriate variable A that characterizes the fan is a fan direction characteristic. Fans have this unique directional feature (A)
Is determined by Also, a set of elements, pixels x, or appropriate transformation variables x ′ in the fan area in a certain direction A is set as a fan set,

[0053]

(13) SetF (A, x or x ') = {x or x' | Fan (A, x)} (13) Fan (A, x) represents an appropriate fan function for deriving a fan set. The (1) linear fan shown in FIG. 6 is the simplest example of a point-symmetric fan frame region. The hatched portion in (1) indicates the area of the linear fan, that is, the fan set. Also, (2) is a conceptual diagram of a generalized fan, and the hatched portion represents a fan area as in (1). The fan direction feature A is calculated from surrounding pixels. There are various methods for this calculation.
In the present invention, any one of them is used. The geometric direction feature of the image has a direction feature even at the orthogonally transformed coordinates, and geometrically forms an orthogonal relationship. A fan set having the characteristic of a point-symmetric geometric figure has the same geometric characteristic even after orthogonal transformation.

Referring back to FIG. The fan pixel correlation evaluation unit 101b calculates the correlation evaluation of an element in the fan area, that is, a fan set element, based on the fan similarity or the fan distance. As an example of fan similarity,
If the set-theoretic similarity is selected, and the reference pixel in the appropriate fan area of the prediction target pixel x_i is x′_i_j, the fan similarity Fm is given by the following equation (3).

[0055]

Fm = (min {x′_i_j}) / (max {x′_i_j}) <= 1 (14)

As in the case of the optimum value calculating means 102, the optimum value calculating means 102b selects a selected set or a suitable optimum function which maximizes the correlation between pixels based on the evaluation value calculated by the fan pixel correlation evaluating means 101b. Then, an optimum value is calculated. In the case of the selection type, the optimum value is expressed as in Expression (5). In the case of the arithmetic type, it is expressed as in equation (6).
In many cases, the optimal function f () included in Equation (6) is the weight w′_i_j of the determined multiple reference pixels x′_i_j.
Can be treated as a linear combination of In this case, the predicted value p_i is

[0057]

(15) p_i = round {[i, j] {w'_i_j * 'x_i_j} (15) The weights w′_i_j are normalized,

[0058]

[I, j] Σw'_i_j = 1 (16)

The optimal DPCM encoding means 113 performs the same processing as the three-dimensional optimal DPCM encoding means 110a.
That is, using the prediction value p_i of Expression (15), encoding is performed according to Expression (11).

[0060]

(17) d_i = x_i-round {[i, j]} w'_i_j * 'x_i_j} (17) Decoding is from equation (12):

[0061]

X_i = d_i + round {[i, j]} w'_i_j * 'x_i_j} (18)

The operation of the image compression apparatus having such a configuration will be described. Three-dimensional fan optimal DPCM encoding means 1
When a three-dimensional digital image signal is input to 10b, correlation evaluation of each pixel that can be referred to is performed by the fan pixel correlation evaluation unit 101b. Fan pixel correlation evaluation means 101b
In a moving image, only pixels in a fan area in the same image are set as reference pixels. The optimum value calculating means 102b calculates the optimum value by, for example, Expression (15). The optimal DPCM encoding unit 113 uses the predicted value of Expression (15) to calculate Expression (1).
DPCM encoding is performed as in 7). Since the entropy of the signal subjected to the DPCM encoding is reduced by the optimal adaptive processing, the entropy encoding by the entropy encoding unit 200 generates a lossless compressed image signal having a high compression rate. Further, the image decompression device performs the same optimal adaptation processing as described above, and obtains DP according to Expression (18).
Perform CM decoding.

The case where the second optimal adaptive coding means 103 is a one-layer subband conversion means will be described. Here, similarly to the first case described above, an optimal one-layer subband conversion method for selecting an arbitrary referenceable pixel and calculating an optimal value, and a fan optimization for restricting the referenceable pixel within the fan area. The one-layer subband conversion method will be described sequentially.

The optimal one-layer subband conversion method will be described. FIG. 7 is a configuration diagram of an image compression apparatus that performs optimal one-layer subband conversion according to an embodiment of the present invention.
1 and 4 are denoted by the same reference numerals, and description thereof is omitted. This image compression apparatus receives a three-dimensional digital image signal and performs one-layer sub-band conversion (hereinafter referred to as sub-band conversion) as a three-dimensional optimum sub-band converter 120a.
Having.

The three-dimensional optimal sub-band converting means 120a
The pixel correlation evaluation unit 101a that evaluates the correlation of each of the referenceable pixels of the prediction target pixel, the optimum value calculation unit 102a that calculates the optimum value based on the pixel correlation evaluation of the pixel correlation evaluation unit 101a, And sub-band conversion means 123 for performing sub-band conversion based on the sub-band conversion.

The pixel correlation evaluating means 101a of the three-dimensional optimal DPCM encoding means 110a
The same processing as in a is performed. That is, referenceable pixels around the prediction pixel in the current reference image to which the prediction pixel belongs,
Further, when no motion is detected around the predicted pixel, the corresponding pixel of the temporal reference image is set as a referenceable pixel, and the correlation of each reference pixel is evaluated. For example, the evaluation of the similarity is performed by Expression (8).

The optimum value calculating means 102a also has a three-dimensional optimum D
The same processing as that performed by the optimum value calculation unit 102a of the PCM encoding unit 110a is performed. That is, an optimal value is calculated by obtaining a selected set or an optimal function that maximizes the correlation between pixels. As described above, when the optimal function is treated as a linear combination of the determined weights of the plurality of reference pixels, the prediction value p_i
Can be represented by equation (9).

The optimum sub-band converting means 123 performs sub-band conversion based on the predicted value p_i calculated by the optimum value calculating means 102a, and divides the three-dimensional image signal into predetermined bands. The band-divided low-frequency component L_i is calculated based on the original pixel x
_2i, x_ (2i + 1),

[0069]

L_i = x_2i + floor (f (L, H) * C) (19) Here, floor () is a function for rounding down a decimal part, and f (L, H) is an appropriate conversion variable. C is an appropriate processing coefficient. In many cases, it can be considered that C = 0, in which case the low frequency component L_i is

[0070]

L_i = x — 2i (20) Further, the high frequency component H_i is calculated using the predicted value p_i,

[0071]

H_i = x_ (2i + 1) −p_i (21) As is clear from equation (19),
If the prediction value p_i can be accurately predicted, H_i
Has a small value, and the entropy can be reduced.

The decoding process of the signal subjected to the sub-band transform coding according to the equations (20) and (21) is as follows.

[0073]

X 22 i = L_i (22)

[0074]

X_ (2i + 1) = H_i + p_i (23)

The operation of such an image compression device will be described. When a three-dimensional digital image signal is input to the three-dimensional optimal sub-band conversion means 120a, the pixel correlation evaluation means 1
The correlation evaluation of each pixel which can be referred to by 01a is performed. As a result of the motion detection, if it is an area that can be regarded as a still image, pixels in the t direction are added as reference pixels, and in an area that can be regarded as a moving image, only predetermined peripheral pixels in the same image are set as reference pixels. Also, after local motion position correction by motion vector calculation, motion detection can be performed again. The optimum value calculation means 102a calculates the optimum value by, for example, Expression (9). The optimum sub-band conversion unit 123 performs sub-band conversion as shown in Expressions (20) and (21), and divides the three-dimensional digital image signal into bands. Since the predicted value p_i has been optimally adaptively processed and the entropy of the high-frequency component H_i obtained from the equation (21) has been reduced, the entropy coding means 200 performs entropy coding to obtain a high compression ratio without a high compression ratio. A lossy compressed image signal is generated.

Generally, the optimum adaptive processing described above is applied to the low-frequency component L_i, and subband conversion is performed, thereby hierarchically band-dividing the original three-dimensional digital image signal. FIG. 8 shows a hierarchical structure of the optimal one-layer subband conversion according to an embodiment of the present invention. The input three-dimensional digital image signal is subjected to optimal one-layer subband conversion according to the procedure described above, and the low-frequency component L1 and the high-frequency component H1
Is divided into bands. Further, the optimum one-layer subband conversion is performed on the low-frequency component L1 according to the procedure described above, and the band is divided into a low-frequency component L2 and a high-frequency component H2. Subsequently, the low-frequency component L2 is similarly band-divided into a low-frequency component L3 and a high-frequency component H3. Furthermore, by performing the optimal one-layer subband conversion of the above-described procedure on the low-frequency components as necessary,
Form a hierarchical structure. FIG. 9 shows an image diagram obtained by the optimal one-layer subband conversion according to the embodiment of the present invention. The same components as those in FIG. 8 are denoted by the same numbers. In the first stage (1), the three-dimensional digital image signal is band-divided into a low-frequency component L1 and a high-frequency component H1 by optimal one-layer subband conversion. In the first stage (1), optimal adaptive subband conversion is performed in the vertical direction, and the low-frequency component L1 is placed on the left and the high-frequency component H1 is placed on the right. In the next second step (2), the low-pass component L1 is subjected to optimal adaptive sub-band conversion in the horizontal direction, and divided into a low-pass component L2 and a high-pass component H2. In the following third stage (3), the low-pass component L2 in the second stage is subjected to optimal adaptive sub-band conversion in the vertical direction, and divided into a low-pass component L3 and a high-pass component H3. As described above, the entropy of the high frequency component is reduced, and the entropy encoding unit 200 generates a three-dimensional lossless compressed image signal at a high compression rate.

Further, the image decompression device performs the same optimal adaptation processing as described above, and performs decoding according to equations (22) and (23). Next, the fan-optimized one-layer subband conversion method will be described. FIG. 10 is a configuration diagram of an image compression device that performs fan-optimized one-layer subband conversion according to an embodiment of the present invention. 1, 5, and 7 are denoted by the same reference numerals, and description thereof will be omitted. This image compression apparatus has a three-dimensional fan optimum sub-band conversion unit 120b that inputs a three-dimensional digital image signal and performs sub-band conversion.

Three-dimensional fan optimum sub-band conversion means 12
0b is a fan pixel correlation evaluation unit 101b that evaluates the correlation of each of the referenceable pixels in the fan area, an optimum value calculation unit 102b that calculates an optimum value based on the pixel correlation evaluation of the fan pixel correlation evaluation unit 101b. , An optimal sub-band conversion unit 123 that performs sub-band conversion based on the optimal value.

The fan pixel correlation evaluation means 101b performs the same processing as the fan pixel correlation evaluation means 101b of the three-dimensional fan optimum DPCM encoding means 110b. That is, elements in the fan area of the current reference image to which the prediction target pixel belongs are referred to as referenceable pixels, and the correlation of each of these pixels is evaluated. For example, if the fan similarity is selected as the evaluation method, the correlation evaluation value can be calculated by Expression (14).

The optimum value calculating means 102b is also the optimum value calculating means 10 of the three-dimensional fan optimum DPCM coding means 110b.
The same processing as in 2b is performed. That is, an optimal value is calculated by obtaining a selected set or an optimal function that maximizes the correlation between pixels. For example, when the optimal function f () is treated as a linear combination of the determined weights w′_i_j of the plurality of reference pixels x′_i_j, the prediction value p_i can be expressed as in Expression (15).

The optimal sub-band converting means 123 performs the same processing as the three-dimensional optimal sub-band converting means 120a.
Here, as represented by Expression (15), assuming that the optimal value obtained by the optimal function is FP {x_ (2n-j)},
Encoding is performed according to equations (20) and (21) using the low-frequency component L_
i and the high frequency component H_i are

[0082]

L_i = x_2i (24)

[0083]

H_i = x_ (2i + 1) −FP {x_ (2n−j)} (25) Equation (24) is the same as equation (20). Here, FP represents all the function conversions that output the fan optimum value. In addition, decoding is performed according to equations (22) and (2).
3) From

[0084]

## EQU26 ## x_2i = L_i (26)

[0085]

X_ (2i + 1) = H_i + FP {x_ (2n−j)} (27) Equation (26) is the same as equation (22).

The operation of such an image compression device will be described. Three-dimensional fan optimal subband converter 120b
When the three-dimensional digital image signal is input to the fan pixel correlation evaluation unit 101b, the correlation evaluation of each of the referenceable pixels in the fan area is performed. Optimal value calculation means 10
2a calculates an optimum value by, for example, equation (15).
The optimum sub-band converting means 123 calculates the equations (24) and (2)
The sub-band conversion is performed as shown in 5), and the three-dimensional digital image signal is divided into bands. The predicted value p_i has been optimally adaptively processed, and the high-frequency component H obtained from Expression (25)
Since the entropy of _i is reduced, a lossless compressed image signal with a high compression rate is generated by entropy encoding by the entropy encoding unit 200. Further, the above-described optimal adaptation processing is further performed on the low-frequency component L_i to perform sub-band conversion, thereby hierarchically band-dividing the original three-dimensional digital image signal. When performing the sub-band conversion by the fan optimum adaptation process, the directional characteristics of the image are calculated from the peripheral pixels of each stage in each layer. Further, since the hierarchical structure stores the basic structure of the image, most of the fan features are also stored. By utilizing this feature, it is also possible to use the fan feature of the higher hierarchy for the fan calculation of the lower hierarchy.

The image decompression device performs the same optimal adaptive processing as described above, and performs decoding according to equations (26) and (27). The case where the third optimal adaptive encoding unit 103 is a multi-layer one-layer subband conversion unit will be described. In the multi-layer one-layer sub-band conversion, depending on whether or not sub-sampling is performed, a reference pixel has an inverted concave structure (no sub-sampling) or a shell-shaped structure. Further, as in the first case described above, the reference pixels are classified according to whether they are restricted to the fan area, and any referenceable pixel is selected to calculate an optimum value. A band method, a multi-layer inverted concave fan optimal single layer sub-band method that limits the pixels that can be referred to the fan area, a multi-layer shell structure optimal single layer sub-band method, and a multi-layer shell structure fan optimal single layer sub-band method, Will be described.

The multi-layer inverted concave structure optimal single layer sub-band method will be described. FIG. 11 is a configuration diagram of an image compression apparatus that performs multi-layer inverted concave structure optimal one-layer subband conversion according to an embodiment of the present invention. The same components as those in FIG. 1 are denoted by the same reference numerals, and description thereof is omitted. This image compression apparatus inputs a three-dimensional digital image signal and performs a three-dimensional inverted concave structure optimal subband conversion unit that performs multi-layer inverted concave structure optimal single layer subband conversion (hereinafter referred to as inverted concave structure optimal subband conversion). 130a.

Three-dimensional inverted concave structure optimal subband conversion means 1
30a is a pixel correlation evaluation means 131 for evaluating the correlation of each reference pixel having an inverted concave structure with respect to the prediction target pixel.
a, an optimum value calculation unit 132a that calculates an optimum value based on the pixel correlation evaluation of the pixel correlation evaluation unit 131a, and an inverted concave structure optimum subband conversion unit 133 that performs multi-layer subband conversion based on the optimum value. Be composed.

The pixel correlation evaluation means 131a evaluates the correlation between each of the predetermined referenceable pixels belonging to the current reference pixel and the temporal reference pixel of the prediction target pixel. Here, a referenceable pixel around the prediction pixel in the current reference image to which the prediction pixel belongs, and a pixel that can refer to a corresponding pixel in the temporal reference image when no motion is detected around the prediction pixel. I do. The current reference image will be described. FIG.
Is an example of a pixel array having an inverted concave structure. In the multi-layer inverted concave subband conversion, a pixel is divided into two stripes. As shown in the figure, a layer to which, for example, a pixel indicated by a black circle belongs is a known layer among the layers of the stripe structure divided into two. Further, pixels that do not belong to the known layer are predicted from pixels of the known layer and a preceding pixel that is decoded first. In the simplest example, the target pixel p is predicted from a pixel of a known layer indicated by a black circle and a reference pixel having an inverted concave structure of a front pixel indicated by a cross. In this way, the pixels that constitute the inverted concave structure around the target pixel p are referenceable pixels.

Returning to FIG. 11, the description will be continued. In the pixel correlation evaluation means 131a, the pixel correlation degree Cm (x
— 2n + d, k, t). d is generalized as an appropriate integer indicating an appropriate sample for the optimal processing in the d-sequence determined from the stripe structure. Also, k is an integer indicating an appropriate pixel determined by the inverted concave structure, and t is an integer indicating an appropriate temporal reference pixel.
In the case of the simplest seven-point inverted-concave structure, k is a pixel indicated by the symbol x shown in FIG. 12, but is generally conveniently selected from all pixels in the past processes. The correlation is, for example, the similarity Sm
in the case of,

[0092]

Sm = (min {x_2n + d, k, t}) / (max {x_2n + d, k, t}) <= 1 (28) In addition, the pixel correlation degree Cm is calculated by any of the methods for evaluating the correlation such as the distance.

The optimum value calculating means 132a calculates an optimum value P which maximizes the correlation Cm. The optimal value P is obtained by a process of selecting a peripheral pixel set that maximizes the correlation,

[0094]

P = round {BP} x_ (2n + d, k, t) {for max {Cm}} (29) Here, BP ｛｝ is an optimal function for calculating an optimal value under a condition of for or less. Round ｛｝ represents the conversion of the function in ｛｝ into an integer.

This processing includes not only the optimum selection but also the continuous change of the processing for each pixel. This is generally calculated as a sum of the optimally referenced neighboring pixel w_p and the optimal weight w_p.

[0096]

BP (w_p, x_p) = [parameter p] Σw_p * x_p (30) Here, the weight w_p is

[0097]

31 is normalized so that [parameter p] ａw_p = 1 (31) There are various methods for calculating w_p, and one of them is used as appropriate.

The inverse concave structure optimum sub-band conversion means 133 performs sub-band conversion based on the calculated optimum value. Subband transform coding is

[0099]

L_n = x_2n + d (32)

[0100]

H_n = x_ (2n−1 + d) −round {BP} x_ (2n + d, k, t) {for max {Cm} (33)

By repeating the processing described above,
The image is divided hierarchically. The multi-layer inverted concave structure type has a structure in which an image is divided into multi-layers by a stripe structure, and an optimum process is performed based on both the other layer information and process information, that is, information that has been previously processed. FIG. 13 shows an encoding structure of a multi-ear inverted concave structure type optimal subband transform. Decomposition into two layers (U, V) by the optimal subband conversion, and continuous for each pixel from three pieces of information (U, V, W) of process information W composed of the current reference pixel and the temporal reference pixel An optimal process for calculating a variable optimal weight coefficient is performed.

The inverse sub-band transform decoding is as follows.

[0103]

X 34 n + d = L_n (34)

[0104]

X_ (2n-1 + d) = H_n + round {BP} x_ (2n + d, k, t)} for max {Cm} (35)

The operation of the image compression apparatus will be described with reference to FIG. When the three-dimensional digital image signal is input to the three-dimensional inverted-concave structure optimal sub-band conversion unit 130a, the pixel correlation evaluation unit 131a evaluates the correlation of each of the referenceable pixels having the inverted-concave structure. Further, as a result of the motion detection, if the area can be regarded as a still image, the pixels of the temporal reference image in the t direction separated from the current reference image by time are added as referenceable pixels. In a region that can be regarded as a moving image, only predetermined peripheral pixels in the same image are set as reference pixels. Also, after local motion position correction by motion vector calculation, motion detection can be performed again. The optimum value calculation unit 132a calculates the optimum value by using, for example, Expression (29) or (30). Inverted concave structure optimal subband converter 1
33 performs subband conversion as represented by equations (32) and (33), and divides the three-dimensional digital image signal into layers. The predicted value has been optimally adaptively processed, and the equation (3)
Since the entropy of the high-frequency component H_n obtained in 3) is reduced, the entropy encoding unit 200 performs entropy encoding to generate a lossless compressed image signal with a high compression ratio.

Further, the image decompression device performs the same optimal adaptation processing as described above, and performs inverse subband conversion according to equations (34) and (35) to decode the target image signal. Next, the multi-layer inverted concave fan optimal single layer sub-band method will be described. FIG. 14 is a configuration diagram of an image compression device that performs multi-layer inverted concave structure fan optimal single layer sub-band conversion according to an embodiment of the present invention. FIG.
The same reference numerals are given to the same components as those in FIG. 11 and the description will be omitted. This image compression device has a three-dimensional inverted concave fan optimum sub-band conversion unit 130b that receives a three-dimensional digital image signal and converts the inverted concave fan optimum sub-band.

The three-dimensional inverted concave fan optimum sub-band converting means 130b is a pixel correlation evaluating means 1 for evaluating the correlation of each reference pixel having an inverted concave structure in the fan area.
31b, an optimum value calculating means 132b for calculating an optimum value based on the pixel correlation evaluation of the pixel correlation evaluation means 131b, and an inverted concave structure optimum subband converting means 133 for performing multi-layer subband conversion based on the optimum value. Be composed.

The fan pixel correlation evaluation means 131b evaluates the correlation between the current reference pixel of the prediction target pixel and the predetermined referenceable pixels belonging to the temporal reference pixel. here,
An element in the fan area of the current reference image to which the prediction pixel belongs is set as a referenceable pixel. It is assumed that the fan area exists only in the current reference image. The fan pixel correlation evaluation unit 131b calculates the correlation evaluation of an element having an inverted concave structure in the fan area, that is, a fan set element, based on the fan similarity or the fan distance. As an example of the fan similarity, if a set-theoretic similarity is selected, a reference pixel in an appropriate fan area of the prediction target pixel x_n is defined as x_2n +
Assuming d and k, the fan similarity Fm is calculated from the equation (14) as follows:

[0109]

Fm = (min {x_2n + d, k}) / (max {2n + d, k}) <= 1 (36)

The optimum value calculating means 132b calculates an optimum value P which maximizes the correlation Fm. The optimal value P is obtained by a process of selecting a peripheral pixel set that maximizes the correlation,

[0111]

P = round {FP} x_ (2n + d, k) {for max} Fm} (37) Here, FP # is an optimal function for calculating an optimal value under the condition of for or less. Round ｛｝ represents the conversion of the function in ｛｝ into an integer.

Inverted concave structure optimal subband conversion means 133
Is the same as the inverse concave structure optimal subband conversion means 133 of the three-dimensional inverted concave structure optimal subband conversion means 130a,
Subband conversion is performed based on the calculated optimum value. Subband transform coding is

[0113]

L_n = x_2n + d (38)

[0114]

H_n = x_ (2n-1 + d) -round {FP} x_ (2n + d, k)} for max {Fm} (39) Equation (38) is the same as equation (32).

The inverse sub-band transform decoding is

[0116]

## EQU40 ## x_2n + d = L_n (40)

[0117]

X_ (2n-1 + d) = H_n + round {FP} x_ (2n + d, k)} for max {Fm} (41) Equation (40) is the same as equation (34).

The operation of such an image compression device will be described. When the three-dimensional digital image signal is input to the three-dimensional inverted concave fan optimum sub-band conversion means 130b, the fan pixel correlation evaluation means 131b evaluates the correlation of each of the referenceable pixels in the fan area. The optimum value calculating means 132a calculates the optimum value by, for example, Expression (37). The inverse concave structure optimal sub-band conversion unit 133 performs sub-band conversion as shown in Expressions (38) and (39), and divides the input image signal into layers. Since the predicted value has been optimally adaptively processed and the entropy of the high-frequency component H_i obtained from the equation (39) has been reduced, the entropy coding unit 200 performs entropy coding to achieve a high compression rate lossless. A compressed image signal is generated. Further, the above-described optimal adaptive processing is further performed on the low-frequency component L_i to perform subband conversion,
The original three-dimensional digital image signal is hierarchically divided into bands.
When performing the sub-band conversion by the fan optimum adaptation process, the directional characteristics of the image are calculated from the peripheral pixels of each stage in each layer. Further, since the hierarchical structure stores the basic structure of the image, most of the fan features are also stored.
By utilizing this feature, it is also possible to use the fan feature of the higher hierarchy for the fan calculation of the lower hierarchy.

Further, the image decompression device performs the same optimal adaptation processing as described above, and performs inverse subband conversion according to equations (40) and (41) to decode the target image signal. Next, the multi-layer shell structure optimal single layer subband method will be described. FIG. 15 is a configuration diagram of an image compression apparatus that performs one-layer subband conversion optimal for a multi-layer shell structure according to an embodiment of the present invention. The same components as those in FIG. 1 are denoted by the same reference numerals, and description thereof is omitted. This image compression apparatus has a three-dimensional shell structure optimal sub-band conversion unit 140a that receives a three-dimensional digital image signal and performs multi-layer shell structure optimal one-layer sub-band conversion (hereinafter referred to as shell structure optimal sub-band conversion). .

[0120] Three-dimensional shell structure optimal subband conversion means 1
Reference numeral 40a denotes a pixel correlation evaluation unit 141 that evaluates the correlation of each reference pixel having a shell structure with respect to the prediction target pixel.
a, an optimum value calculating means 142a for calculating an optimum value based on the pixel correlation evaluation of the pixel correlation evaluating means 141a, and a shell structure optimum subband converting means 143 for performing multi-layer subband conversion based on the optimum value. Is done.

The pixel correlation evaluation means 141a evaluates the correlation between the current reference pixel of the prediction target pixel and the predetermined referenceable pixels belonging to the temporal reference pixel. Here, the referenceable pixels around the prediction pixel in the current reference image to which the prediction pixel belongs, and further, if no motion is detected around the prediction pixel, the corresponding pixel of the temporal reference image is referred to as the referenceable pixel. And The pixels of the current reference image will be described. FIG. 16 is an example of a pixel array having a shell structure. In the multi-layer shell structure sub-band conversion, pixels are alternately divided into two. As shown in the figure, a layer composed of pixel numbers x_ (2i + d) represented by double circles and a layer composed of pixel numbers x_ (2i-1 + d) represented by single circles Is divided into two. For example,
If x_ (2i + d) is a known layer, the target pixel p
Pixel a belonging to a known layer indicated by a black circle around
The target pixel p can be predicted using b, c, and d as reference pixels. Such a structure of the reference pixel is a diamond structure. Further, the front pixels f and g belonging to the same layer as the target pixel p and decoded earlier can be added to the reference pixels. Such a structure of the reference pixel is referred to as a shell structure. As described above, in the simplest example, the target pixel p is predicted from the reference pixel having a shell structure including the pixels of the known layer indicated by black circles and the front pixels indicated by crosses. Since the diamond structure is a kind of the shell structure, the case of the shell structure will be described below.

In the pixel correlation evaluation means 141a, the pixel correlation degree C′m (x_2n + d ′, k ′, k ′) of the referenceable peripheral pixel and the temporal reference pixel m set in the shell structure described above is set.
Find t). d 'has been generalized as a suitable integer indicating a suitable sample for optimal processing in the d' series determined from the diamond structure. Also, k '
Is an integer indicating an appropriate pixel determined by the shell structure, and t is an integer indicating an appropriate temporal reference pixel. k '
Represents the case of the simplest six-point shell structure, as shown in FIG.
The pixel of the mark is generally conveniently selected from all pixels in the past process. Returning to FIG. The correlation is, for example, in the case of the similarity S′m,

[0123]

S′m = (min {x_2n + d ′, k ′, t}) / (max {x_2n + d ′, k ′, t}) <= 1 (42) In addition, the pixel correlation degree C′m is calculated by any of the methods for evaluating the correlation such as the distance.

In the optimum value calculating means 142a, the correlation C'm
The optimal value P ′ that maximizes is calculated. The optimal value P ′ is obtained by a process of selecting a peripheral pixel set that maximizes the correlation,

[0125]

P ′ = round {BP} x_ (2n + d ′, k ′, t) {for max} C′m} (43) Here, BP ｛｝ is an optimal function for calculating an optimal value under a condition of for or less. Round ｛｝ represents the conversion of the function in ｛｝ into an integer.

The shell structure optimum sub-band conversion means 143 performs sub-band conversion based on the calculated optimum value. Subband transform coding is

[0127]

L_n = x_2n + d '(44)

[0128]

H_n = x_ (2n−1 + d ′) − round {BP} x_ (2n + d ′, k ′, t) {for max {C′m} (45)

By repeating the processing described above,
The image is divided hierarchically. Also, the inverse subband transform decoding

[0130]

## EQU46 ## x_2n + d '= L_n (46)

[0131]

X_ (2n-1 + d ') = H_n + round {BP} x_ (2n + d', k ', t) {for max {C'm} (47)

The operation of such an image compression device will be described. Three-dimensional shell structure optimal subband conversion means 140
When a three-dimensional digital image signal is input to “a”, the pixel correlation evaluation unit 141a evaluates the correlation of each of the referenceable pixels having the shell structure. Furthermore, as a result of motion detection,
In a region that can be regarded as a still image, a pixel of the temporal reference image in the t direction separated from the current reference image by a time is added as a referenceable pixel. In a region that can be regarded as a moving image, only predetermined peripheral pixels in the same image are set as reference pixels. Also, after local motion position correction by motion vector calculation, motion detection can be performed again. The optimum value calculation means 142a calculates the optimum value by, for example, Expression (43). The shell structure optimum sub-band conversion means 143 is calculated by the equations (44) and (45).
The sub-band conversion is performed as shown in (1) to divide the three-dimensional digital image signal into predetermined layers. Since the predicted value has been optimally adaptively processed and the entropy of the high-frequency component H_n obtained from Expression (45) has been reduced, the entropy encoding unit 200 performs entropy encoding to achieve a high compression rate lossless. A compressed image signal is generated.

The image decompression device performs the same optimal adaptation processing as described above, and performs inverse subband conversion according to equations (46) and (47) to decode the target image signal. Next, a multi-layer shell structure fan optimal single layer subband method will be described. FIG. 17 is a configuration diagram of an image compression apparatus that performs multi-layer shell structure fan optimal single-layer subband conversion according to an embodiment of the present invention. FIG.
The same components as those in FIG. 15 are denoted by the same reference numerals, and description thereof is omitted. This image compression apparatus has a three-dimensional shell structure fan optimum sub-band conversion unit 140b that inputs a three-dimensional digital image signal and converts the shell structure fan optimum sub-band.

The three-dimensional shell structure fan optimum subband conversion means 140b includes fan pixel correlation evaluation means 141b for evaluating the correlation between individual reference pixels having a shell structure in the fan area, and pixels of the fan pixel correlation evaluation means 141b. Optimum value calculation means 1 for calculating an optimum value based on the correlation evaluation
42b, a shell-structure optimum sub-band conversion unit 143 that performs multi-layer sub-band conversion based on the optimum value,
Consists of

The fan pixel correlation evaluation means 141b evaluates the correlation between the current reference pixel of the prediction target pixel and the predetermined referenceable pixels belonging to the temporal reference pixel. here,
An element in the fan area of the current reference image to which the prediction pixel belongs is set as a referenceable pixel. It is assumed that the fan area exists only in the current reference image. The fan pixel correlation evaluation unit 141b calculates the correlation evaluation of an element having a shell structure in the fan area, that is, the fan set element, based on the fan similarity or the fan distance. As an example of the fan similarity, if a set-theoretic similarity is selected, the reference pixel having a shell structure in an appropriate fan region of the prediction target pixel x_n is x_2n + d ′, k ′, and the fan similarity F
m is given by equation (14).

[0136]

Fm = (min {x_2n + d ', k'}) / (max {2n + d ', k'}) <= 1 (48)

The optimum value calculating means 142b calculates an optimum value P 'that maximizes the correlation Fm. The optimal value P ′ is obtained by a process of selecting a peripheral pixel set that maximizes the correlation,

[0138]

P ′ = round {FP} x_ (2n + d ′, k ′)} for max {Fm} (49) Here, FP # is an optimal function for calculating an optimal value under the condition of for or less. Round ｛｝ represents the conversion of the function in ｛｝ into an integer.

Shell structure optimum sub-band conversion means 143
Is the same as the shell structure optimum sub-band conversion means 143 of the three-dimensional shell structure optimum sub-band conversion means 140a,
Subband conversion is performed based on the calculated optimum value. Subband transform coding is

[0140]

L_n = x_2n + d '(50)

[0141]

H_n = x_ (2n−1 + d ′) − round {FP} x_ (2n + d ′, k ′) {for max {Fm} (51) Equation (50) becomes the same as equation (44).

The inverse sub-band transform decoding is as follows.

[0143]

X 52 n + d ′ = L_n (52)

[0144]

X_ (2n-1 + d ') = H_n + round {FP} x_ (2n + d', k ')} for max {Fm} (53) Equation (53) becomes the same as equation (45).

An operation of such an image compression device will be described. When the three-dimensional digital image signal is input to the three-dimensional shell structure fan optimum sub-band conversion unit 140b, the fan pixel correlation evaluation unit 141b performs correlation evaluation of each of the referenceable pixels in the fan area. The optimum value calculating means 142a calculates the optimum value by, for example, Expression (49). The shell structure optimal sub-band conversion unit 143 performs sub-band conversion as shown in Expressions (50) and (51), and divides the input image signal into layers. Since the predicted value has been optimally adaptively processed and the entropy of the high-frequency component H_n obtained by the equation (51) has been reduced, the entropy encoding unit 200 performs entropy encoding to achieve a high compression rate lossless. A compressed image signal is generated. Further, by applying the above-described optimal adaptive processing to the low-frequency component L_n and performing sub-band conversion,
The original three-dimensional digital image signal is hierarchically divided into bands.
When performing the sub-band conversion by the fan optimum adaptation process, the directional characteristics of the image are calculated from the peripheral pixels of each stage in each layer. Further, since the hierarchical structure stores the basic structure of the image, most of the fan features are also stored.
By utilizing this feature, it is also possible to use the fan feature of the higher hierarchy for the fan calculation of the lower hierarchy.

Further, the image decompression device performs the same optimal adaptation processing as described above, and performs inverse subband conversion according to equations (52) and (53) to decode the target image signal. Note that the above processing functions can be realized by a computer. In this case, the processing contents of the functions that the image compression apparatus and the image expansion apparatus should have are described in a program recorded on a computer-readable recording medium. Then, by executing this program on a computer, the above processing is realized on the computer. Examples of the computer-readable recording medium include a magnetic recording device and a semiconductor memory. When distributing in the market, CD-ROM (Compact Disc Read Only Memory)
The program may be stored and distributed in a portable recording medium such as a disk or a floppy disk, or stored in a storage device of a computer connected via a network, and transferred to another computer via the network. When the program is executed by the computer, the program is stored in a hard disk device or the like in the computer, loaded into the main memory, and executed.

[0147]

As described above, according to the present invention, a three-dimensional digital image signal is input, the correlation of each of the predetermined referenceable pixels of the prediction target pixel is evaluated, and the pixel which maximizes the correlation between the pixels is estimated. An optimal value is calculated by obtaining a selected set of or an appropriate optimal function. The target pixel is coded based on this optimum value, and further subjected to entropy coding to generate a three-dimensional lossless compressed image signal.

As described above, since the optimum adaptive processing based on the optimum value that maximizes the correlation is performed, the prediction accuracy of the predicted value can be improved. By increasing the prediction accuracy,
The entropy of the encoded signal after the encoding process is reduced, and a lossless image compression signal with a higher compression rate can be generated during entropy encoding. Further, the fact that there is no need to add a special code for adaptive processing also contributes to an improvement in the compression ratio.

Further, the image decompression device of the present invention inputs a three-dimensional compressed image signal compressed by the image compression device and performs entropy decoding. In addition, the correlation of each of the predetermined referenceable pixels of the target pixel is evaluated, the optimal value is calculated by calculating a selected set of pixels or an appropriate optimal function that maximizes the correlation between the pixels, and based on the optimal value, Decode the target pixel.

As described above, since the optimum value calculated by the same processing as that of the image compression apparatus is the same as the value at the time of image compression, even if a special code for adaptive processing is not added,
It can be calculated by the image decompression device alone. For this reason,
As a result, it contributes to the improvement of the compression ratio.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of an image compression device according to an embodiment of the present invention.

FIG. 2 shows a pixel correspondence of a three-dimensional image in the image compression apparatus according to the embodiment of the present invention.

FIG. 3 is a configuration diagram of an image decompression device according to an embodiment of the present invention.

FIG. 4 is a configuration diagram of an image compression apparatus that performs optimal DPCM according to an embodiment of the present invention.

FIG. 5 is a fan optimum DPC according to an embodiment of the present invention.
1 is a configuration diagram of an image compression device that performs M.

FIG. 6 is a conceptual diagram of a fan area.

FIG. 7 is a configuration diagram of an image compression device that performs optimal one-layer subband conversion according to an embodiment of the present invention.

FIG. 8 shows a hierarchical structure of optimal one-layer subband conversion according to an embodiment of the present invention.

FIG. 9 is a diagram showing an image obtained by an optimal one-layer subband conversion according to an embodiment of the present invention.

FIG. 10 is a configuration diagram of an image compression apparatus that performs fan-optimized one-layer subband conversion according to an embodiment of the present invention.

FIG. 11 is a configuration diagram of an image compression apparatus that performs multi-layer inverted concave structure optimal one-layer subband conversion according to an embodiment of the present invention.

FIG. 12 is an example of a pixel array having an inverted concave structure.

FIG. 13 shows an encoding structure of a multi-ear inverted concave structure type optimal subband transform.

FIG. 14 is a configuration diagram of an image compression device that performs multi-layer inverted concave structure fan optimal single layer sub-band conversion according to an embodiment of the present invention.

FIG. 15 is a configuration diagram of an image compression apparatus that performs one-layer subband conversion optimal for a multi-layer shell structure according to an embodiment of the present invention.

FIG. 16 is an example of a pixel array having a shell structure.

FIG. 17 is a configuration diagram of an image compression apparatus that performs multi-layer shell structure fan optimal single-layer subband conversion according to an embodiment of the present invention.

FIG. 18 is a configuration diagram of a conventional image compression apparatus that performs a three-dimensional prediction type DPCM process.

FIG. 19 shows a pixel correspondence in a three-dimensional image.

[Explanation of symbols]

100: three-dimensional optimal adaptive encoding means, 101: pixel correlation evaluation means, 102: optimal value calculating means, 103: optimal adaptive encoding means, 200: entropy encoding means, 300
... three-dimensional optimal adaptive decoding means, 301 ... pixel correlation evaluation means, 302 ... optimal value calculating means, 303 ... optimal adaptive decoding means, 400 ... entropy decoding means

 ──────────────────────────────────────────────────続 き Continued on the front page F term (reference) 5C059 KK01 MA05 MA41 MA45 ME02 ME11 NN01 NN24 NN29 PP01 PP04 SS01 SS06 SS12 SS26 TA17 TA21 TA62 TA66 TB04 TC03 TC06 TC13 TD02 UA02 UA05 5J064 AA02 BA04 BA09 BB03 BB12 BC02 BC27 BC27

Claims (15)

[Claims] An image compression apparatus for performing lossless image compression of a three-dimensional image composed of a plurality of screens having a temporal correlation, such as a still image continuously shot image and a moving image, comprising: A pixel correlation evaluation unit that evaluates a correlation between each of the predetermined referenceable pixels belonging to the current reference pixel and the temporal reference pixel of the prediction target pixel, and maximizes a correlation between the pixels based on the pixel correlation evaluation. Optimal value calculating means for calculating an optimal value, optimal adaptive encoding means for encoding a target pixel based on the optimal value and generating an encoded signal, entropy encoding means for entropy encoding the encoded signal An image compression device, comprising: 2. The method according to claim 1, wherein the pixel correlation evaluation unit performs motion detection around the prediction target pixel, and adds the temporal reference pixel to the referenceable pixel when there is no motion. An image compression device as described in the above. 3. The image compression apparatus according to claim 1, wherein the pixel correlation evaluation unit evaluates a correlation between individual elements in a predetermined geometric structure area determined by a directional feature of the image. 4. The predetermined geometric structure area of the pixel correlation evaluation means is a geometric structure area having an arbitrary point-symmetric shape, and the geometric structure areas do not interfere with each other. Image compression device. 5. The image compression apparatus according to claim 1, wherein the optimum value calculating means selects an element set that maximizes the correlation between the pixels, and sets the selected set as an optimum value. 6. The image compression apparatus according to claim 5, wherein said optimum value calculating means further calculates an optimum function for optimizing said selected set. 7. The image compression apparatus according to claim 6, wherein said optimum value calculating means obtains the optimum function from the representative correlation of the neighboring pixels suitable for the selected set of weighting coefficients. 8. The image compression apparatus according to claim 6, wherein said optimum value calculating means further limits said optimum function by a predetermined attribute. 9. The image compression apparatus according to claim 1, wherein said optimum adaptive coding means performs DPCM coding on the target pixel based on the optimum value. 10. The image compression apparatus according to claim 1, wherein said optimal adaptive encoding means performs subband conversion for dividing said three-dimensional image signal into predetermined bands based on said optimal value. 11. The optimal adaptive encoding unit calculates a difference between a prediction value obtained by predicting the three-dimensional image signal from a known layer and a referenceable pixel that does not belong to the known layer based on the optimal value and a prediction target pixel 2. The image compression apparatus according to claim 1, wherein a multi-layer sub-band conversion for dividing the image into a plurality of layers is performed. 12. The pixel correlation evaluation unit evaluates a correlation of each pixel as a referenceable pixel with a pixel belonging to the known layer, a front pixel not belonging to the known layer, and a temporal reference pixel. The image compression device according to claim 11, wherein: 13. The image compression apparatus according to claim 12, wherein pixels belonging to the known layer and front pixel pixels not belonging to the known layer form an inverted concave structure surrounding the prediction target pixel. 14. The image compression apparatus according to claim 12, wherein the pixels belonging to the known layer and the front pixels not belonging to the known layer form a shell-shaped structure surrounding the prediction target pixel. 15. An image decompression device for decompressing a compressed image signal obtained by losslessly compressing a three-dimensional image composed of a plurality of screens having a temporal correlation such as a still image continuously shot image and a moving image. Entropy decoding means for decoding the image compression signal which has been entropy-encoded after encoding; Means, an optimum value calculating means for calculating an optimum value for maximizing a correlation between pixels based on the pixel correlation evaluation, and an optimum adaptation for decoding a target pixel from an encoded signal based on the optimum value. An image decompression device comprising: decoding means.