JPH04260290A - Picture data compression method and picture data compression device and picture data decoder - Google Patents

Abstract

PURPOSE:To obtain a coded data not including redundant information with respect to the picture data compression method and picture data compression device and picture data decoder. CONSTITUTION:Orthogonal transformation is applied to each block comprising MXN picture elements of a picture data representing a multi-valued picture to obtain a coefficient matrix of M rows and N columns comprising conversion coefficients, plural components are extracted from a coefficient matrix corresponding to each block and an extracted coefficient matrix of M rows and N columns including only the component extracted as the effective coefficient having values other than zero is generated. The extracted coefficient matrix is subjected to inverse orthogonal transformation and a picture of each block is decoded and plural picture elements are extracted respectively from the obtained picture, a reduced picture comprising the picture elements is generated corresponding to each block and sets of the reduced pictures corresponding to each block are used as picture data and they are coded and compressed furthermore.

Description

[Detailed description of the invention]

0001

[Industrial Application Field] The present invention relates to an image data compression method and an image data compression apparatus for transform-coding and compressing image data representing a multivalued image, and an image data compression method and apparatus for converting and compressing image data representing a multivalued image. The present invention relates to an image data restoration device that restores data.

[0002]An adaptive discrete cosine transform coding method using two-dimensional orthogonal transform is used as a coding method for compressing the amount of data of multivalued images such as halftone images and color images without impairing their characteristics.
e Cosine Transform, hereinafter referred to as ADCT
method) is widely used. This ADCT method converts a multivalued image into a predetermined number of pixels (for example, 8×8
For each block, the image data is orthogonally transformed to obtain a matrix of transform coefficients (hereinafter referred to as DCT coefficients), and each component of this matrix is set to a corresponding visual adaptation threshold (described later). The amount of data is compressed by quantizing the data using the following method and then variable-length encoding.

0003

2. Description of the Related Art FIG. 13 shows the configuration of an image data compression apparatus to which a conventional ADCT method is applied. Also, in Figure 14,
An example of blocks obtained by dividing a multivalued image is shown. D.C.
The T-transform unit 611 performs two-dimensional discrete cosine transform (hereinafter referred to as DCT transform) processing on the input block, and transforms the input block into 8
It is converted into a matrix with 8 rows and 8 columns (hereinafter referred to as DCT coefficient D). FIG. 15 shows an example of this DCT coefficient D.

Each component of this DCT coefficient D is sequentially inputted to a divider 622 as a dividend by a DCT coefficient input section 621 of a linear quantization section 620. Further, the linear quantization unit 620 is provided with a quantization matrix VTH consisting of visually adaptive thresholds corresponding to each frequency component in advance, and is inputted to this quantization matrix VTH and the quantization matrix VTH prior to encoding one screen. An 8-by-8 matrix consisting of the product with the quantization parameter is stored in the quantization threshold holding unit 623 as the quantization threshold QTH. The above-mentioned visual adaptation threshold is predetermined based on experimental results regarding visual sensitivity to each spatial frequency component, and the quantization control parameter is a coefficient that determines the quantization accuracy of an image.

Therefore, in accordance with the input of each component of the DCT coefficient D, the corresponding component of the above-mentioned quantization threshold QTH is input to the divider 622 as a divisor, so that each component of the DCT coefficient D is visually adapted. A quantized coefficient is obtained by quantizing using a threshold value. Generally, human vision is
It has high sensitivity to low spatial frequencies and low sensitivity to high spatial frequencies. Accordingly, as shown in FIG. 16, the value of each component of the above-mentioned quantization matrix VTH is such that the absolute value of the component corresponding to a low spatial frequency is small, and conversely, the absolute value of the component corresponding to a high spatial frequency is small. The value is set large. Therefore, 8
A matrix with 8 rows and 8 columns (hereinafter referred to as quantization coefficient DQU) is
As shown in FIG. 17, only the component in the upper left corner of the matrix indicating the DC component (hereinafter referred to as the DC component) and a very small number of AC components indicating low spatial frequency components around this DC component are non-zero. Most AC components are effective coefficients that have a value, and most AC components are invalid coefficients that have a value of zero.

The quantization coefficient DQU obtained in this way
is converted into a one-dimensional array using a scanning order called zigza scan shown in FIG. 18, and then input to the encoding unit 631. In addition, this encoding unit 631
is the one-dimensional array described above with the effective coefficient (index) and the continuous length (run) of the invalid coefficients before this index.
Based on the code table 632, each combination is replaced with a code corresponding to its frequency of appearance, variable length coding is performed, and the resulting coded data is sent out.

[0007] Furthermore, the encoded data obtained in this manner is restored to image data by an image data restoration device shown in FIG. Decoding unit 711 of image data restoration device
Contrary to the code table 632 described above, the system has a decoding table 712 that shows combinations of runs and indexes corresponding to codes, and decodes sequentially input codes to find combinations of indexes and runs. , based on these, the quantization coefficient DQU is restored and input to the inverse quantization unit 720.

[0008] This inverse quantization section 720 converts the quantization coefficient DQ
By multiplying each component of U by the corresponding component of the quantization threshold QTH, each component of the quantization coefficient DQU is dequantized and the DCT coefficient D is restored. This DCT coefficient D is the inverse DC
is input to the T transform section 731, and this inverse DCT transform section 731
The image data of the corresponding block is restored by the inverse DCT transformation process.

Normally, as described above, all the effective coefficients included in the quantization coefficient DQU corresponding to each block are encoded and sent as an index, and the quantization coefficient DQU is restored from this encoded data. The image data is restored based on this quantization coefficient DQU. On the other hand, in the case of hierarchical restoration in which a detailed image is restored step by step from a rough image, the above-mentioned encoding unit 631 sequentially extracts some components corresponding to low spatial frequencies of the quantization coefficient DQU, A matrix containing only the extracted components as effective coefficients is similarly variable-length coded and is sent out as coded data. For example, in the first layer of layer restoration, the DC component (hereinafter referred to as DC
Then, in the second and third layers, the two adjacent DC components are extracted.
AC components (indicated by symbols 2 and 3 in Figure 18)
and the three AC components adjacent to these components (Figure 18
(indicated by symbols 4, 3, and 6) are extracted, encoded, and sent out.

Accordingly, in the first layer of hierarchical restoration, the above-mentioned decoding section 711 and inverse quantization section 720 restore the DCT coefficient D in which only the DC component is an effective coefficient. Similarly, in the second layer of layer restoration, only the two AC components described above are effective coefficients.
The T coefficient D is restored, and in the third layer of hierarchical restoration, the DCT coefficient D whose only effective coefficients are the three AC components described above is restored.

[0011] Here, the DC component of the DCT coefficient D corresponds to the DC component of the image data, that is, the average value of the image data, and the AC components included in the first row and the first column correspond to the respective blocks. It corresponds to each spatial frequency component indicating changes in pixel data in the row direction and column direction. Figure 20
shows the basis vector representing the fundamental waveform of each spatial frequency component. In the figure, the 0th order basis vector is a basis vector corresponding to the DC component, and the 1st to 7th order basis vectors are basis vectors corresponding to each component in the first row and first column.

As shown in FIG. 20, since the zero-order basis vector corresponding to the DC component is a constant, by performing inverse DCT transformation on the DCT coefficient D of the first layer described above,
An image of one block is restored in which all pixel data has the average value of the pixel data of each block, and a rough image can be obtained from these blocks. Also,
The first-order basis vector corresponding to the AC component included in the DCT coefficient D restored in the second layer described above is:
It shows a monotonous decrease (or increase) in the row or column direction. Therefore, the above-mentioned DCT coefficient D is inverse DCT
By performing the conversion, one block of image consisting of pixels whose pixel data monotonically decreases (or increases) in the row and column directions is obtained.

Similarly, by inverse DCT-transforming the DCT coefficients D restored in the third layer, image data in which pixel data changes according to the second-order basis vector is obtained. In this way, by superimposing the image data sequentially restored in each layer on the above-mentioned rough image, a more detailed image is gradually restored.

[0014]

[Problem to be Solved by the Invention] By the way, when the DCT coefficient D includes only a DC component and three AC components surrounding this DC component as effective coefficients,
One block of image obtained by inverse DCT transformation is an image in which the value of each pixel monotonically decreases (or monotonically increases) according to the first-order base vector. That is, in this case,
Once the pixel data corresponding to the pixels at the four corners of one block are obtained, the values of other pixel data can be approximated by using these pixels as base points and interpolating using the waveform of the first-order basis vector. be able to. Furthermore, if the DCT coefficient D includes a component corresponding to the quadratic basis vector as an effective coefficient, the waveform representing the change in pixel data in the restored image has one extreme value in the row and column directions. expected to have. Therefore, in this case, the pixel that takes the above-mentioned extreme value is set as the inflection point, and the pixel data corresponding to other pixels is approximately determined from the pixel data of the pixel corresponding to this inflection point and the above-mentioned base point. Can be done.

[0015] Therefore, in applications where it is permissible to obtain pixel data approximately as described above, it is sufficient to transmit only information regarding the base point and the inflection point, and information such as the quantization coefficient DQU is sufficient. Furthermore, it is not necessary to transmit information regarding all pixels of one block as encoded data. However, in the conventional system, the DCT coefficients D are encoded and transmitted for each of a huge number of blocks, regardless of whether the above-described approximation is allowed or not. For this reason, the encoded data includes redundant information regarding pixels other than the base point and inflection point for each block, which increases the amount of encoded data as a whole and increases the time required to transmit the encoded data. It was long.

On the other hand, in the lower layers of hierarchical restoration, there is a desire to restore and output a rough image as quickly as possible. There is a need for an image data compression method that encodes at a compression rate and a corresponding image data restoration method. SUMMARY OF THE INVENTION An object of the present invention is to provide an image data compression method and an image data compression device that obtain encoded data that does not include redundant information, and an image data restoration method and image data restoration device that restore this encoded data.

[0017]

[Means for Solving the Problem] FIG. 1(a) is a diagram showing the principle of the invention according to claim 2. The invention of claim 1 orthogonally transforms image data representing a multivalued image for each block of M×N pixels to obtain a coefficient matrix of M rows and N columns each consisting of transform coefficients, and calculates the coefficients corresponding to each block. A plurality of components are each extracted from the matrix, an M-by-N extraction coefficient matrix containing only the extracted components as effective coefficients having non-zero values is generated, and the extraction coefficient matrix is inversely orthogonally transformed, and each block is The image is restored, multiple pixels are extracted from the resulting image, a reduced image consisting of these pixels is generated for each block, and a set of reduced images corresponding to each block is created as image data. , is characterized by being re-encoded and compressed.

FIG. 1(b) is a diagram showing the principle of the invention according to claim 2. The invention according to claim 2 provides a method for obtaining a coefficient matrix by orthogonally transforming image data representing a multivalued image for each block;
Analyze the features of the multivalued image based on the coefficient matrix corresponding to each block, extract components corresponding to the features of the multivalued image from the coefficient matrix corresponding to each block, generate an extraction coefficient matrix, and extract The coefficient matrix is inversely orthogonally transformed to restore the image of each block, each pixel corresponding to the feature of the multivalued image is extracted from the obtained image, and a reduced image consisting of these pixels is created corresponding to each block. The feature is that a set of reduced images corresponding to each block is generated and re-encoded and compressed as image data.

FIG. 2 is a diagram showing the configuration of an image data compression apparatus according to the third aspect of the invention. The invention according to claim 3 provides an orthogonal transformation means 111 for orthogonally transforming image data representing a multivalued image block by block to obtain a coefficient matrix, and extracting a plurality of components from the coefficient matrix corresponding to each block, and extracting each component from the coefficient matrix corresponding to each block. Extraction means 11 for generating an extraction coefficient matrix corresponding to a block
2, an inverse orthogonal transform means 113 that restores the image of each block by inversely orthogonally transforming the extraction coefficient matrix, and a plurality of components extracted by the extraction means 112 from the image data obtained by the inverse orthogonal transform means 113. a reduced image generating means 114 that extracts pixels corresponding to the pixels and generates a reduced image consisting of these pixels corresponding to each block, and a conversion code that converts and encodes image data consisting of the reduced image corresponding to each block. The present invention is characterized in that it includes a converting means 115.

According to a fourth aspect of the present invention, in the image data compression apparatus according to the third aspect, the extraction means 112 extracts a plurality of components including a DC component of the coefficient matrix and an AC component around it, and extracts an extracted coefficient matrix. It is characterized by a configuration that generates. According to a fifth aspect of the invention, in the image data compression apparatus according to the third aspect, the reduced image generating means 114 extracts based on base vectors corresponding to a plurality of components included as effective coefficients in the extraction coefficient matrix. It is characterized by a configuration that determines pixels.

According to a sixth aspect of the present invention, in the image data compression apparatus according to the third aspect, a feature analysis means 1 for analyzing features of a multivalued image based on a coefficient matrix corresponding to each block.
21, and the extraction means 112 is characterized in that it is configured to determine the components to be extracted from the coefficient matrix according to the analysis result by the feature analysis means 121. FIG. 3 shows claim 7.
1 is a diagram showing the configuration of an image data compression device according to the invention; FIG.

According to a seventh aspect of the present invention, in the image data compression apparatus according to the third aspect, in place of the transform encoding means 115, a predictive encoding means for predictively encoding image data consisting of a reduced image corresponding to each block is provided. 131. FIG. 2 is a diagram showing the configuration of an image data compression device according to an eighth aspect of the invention. The invention of claim 8 is defined as claim 3.
In the image data compression device described above, the extraction means 112 quantizes the components of the input coefficient matrix using the corresponding quantization threshold to obtain quantization coefficients, and the extraction means 141 extracts the quantization coefficients from the input quantization coefficients. The present invention is characterized in that it includes an inverse quantization means 142 that performs inverse quantization using a corresponding quantization threshold, and generates an extraction coefficient matrix consisting of transform coefficients obtained by inversely quantizing quantized coefficients.

FIG. 4 is a diagram showing the configuration of an image data restoring apparatus according to the ninth aspect of the invention. The invention of claim 9 provides reduced image restoring means 151 for restoring a set of reduced images based on input encoded data; and reduced image restoring means 151.
a block restoring means 152 for restoring the corresponding block by determining pixel data of other pixels forming the corresponding block based on the pixel data of the plurality of pixels forming each reduced image obtained by It is characterized by

[0024]

[Operation] The invention of claim 1 restores an image represented by a superposition of spatial frequencies corresponding to the extracted components by performing inverse orthogonal transformation on the extracted coefficient matrix consisting of components extracted from the coefficient matrix. By re-encoding a set of reduced images obtained by extracting a plurality of pixels from , as image data, it is possible to obtain encoded data that does not include redundant information about pixels that can be approximately determined.

The invention of claim 2 extracts components corresponding to the features of the multivalued image from the coefficient matrix by analyzing the features of the multivalued image based on the coefficient matrix of each block, and extracts components corresponding to the features of the multivalued image from the reduced image. Pixels corresponding to the features of a multivalued image can be extracted. The invention of claim 3 provides extraction means 1
12 extracts a plurality of components from the coefficient matrix obtained by the orthogonal transform means 111 to generate an extracted coefficient matrix, and the inverse orthogonal transform means 113 performs inverse orthogonal transform on this extracted coefficient matrix to correspond to each block. Get the image you want. Based on this image, the reduced image generation means 114 generates a reduced image corresponding to each block, and the transformation encoding means 115 transforms and encodes a set of these reduced images, so that it can be approximately determined. It is possible to obtain encoded data that does not include redundant data regarding pixels that can be processed.

In the fourth aspect of the invention, the extraction means 112 extracts a plurality of components including a DC component and surrounding AC components, and the inverse orthogonal transformation means 113 roughly approximates the gradation in the original image. An image that reflects the changes is obtained. This makes it possible to eliminate redundant data and obtain a high compression rate, as well as to obtain encoded data that includes information regarding approximate gradation changes in the original image.

The invention of claim 5 provides reduced image generation means 11
4 determines the pixels to be extracted according to the basis vectors corresponding to the effective coefficients included in the extraction coefficient matrix.
Pixels that characterize tone changes in the original image can be extracted from the restored image. This makes it possible to eliminate redundant data and obtain a high compression rate, as well as to obtain encoded data that efficiently includes information regarding approximate gradation changes in the original image.

According to the sixth aspect of the invention, the feature analysis means 121 analyzes the features of the multi-valued image, and the extraction means 112 determines the components to be extracted according to the analysis results by the feature analysis means 121, so that the multi-valued image is It is possible to generate an extraction coefficient matrix that includes components corresponding to image features as effective coefficients. As a result, it is possible to handle multivalued images with various characteristics.
Encoded data that does not include redundant data can be obtained.

[0029] The invention of claim 7 provides a predictive encoding means 131.
By using , information regarding the reduced image is encoded without being lost, so redundant data can be eliminated and deterioration in image quality can be suppressed. In the invention of claim 8, the transform coefficients are once quantized by the quantization means 141 and the dequantization means 142, and then dequantized, and the quantized coefficients are dequantized and obtained by the extraction means 112. An extraction coefficient matrix consisting of transform coefficients can be obtained, thereby making it possible to eliminate noise components contained in the original transform coefficients.

[0030] The invention of claim 9 provides a reduced image restoring means 15.
1 restores a set of reduced images based on encoded data, and the block restoration means 152 restores corresponding blocks based on each reduced image, thereby restoring image data.

[0031]

Embodiments Hereinafter, embodiments of the present invention will be described in detail based on the drawings. FIG. 5 shows the configuration of an embodiment of the image data compression device of the present invention. In FIG. 5, image data input to the image data compression device is subjected to DCT transformation for each block of 8×8 pixels by a DCT transformation unit 211 corresponding to the orthogonal transformation means 111, and sequentially transferred to a buffer 22.
It is held at 1.

In addition, in FIG. 5, the buffer 2 described above
21, the address calculation section 231, the readout circuit 232, and the buffer 233 form the extraction means 112. In this extraction means 112, an address calculation unit 231 sequentially calculates the addresses of the buffer 221 corresponding to predetermined components included in the DCT coefficient D of each block, and
32 to read out the corresponding component. This address calculation unit 231, for example,
The DC component and the three surrounding AC components may be used as the above-mentioned predetermined components, and the addresses corresponding to these components of the DCT coefficient D of each block may be calculated. Further, the readout circuit 232 is configured to store the read component in response to the above-mentioned instruction at the corresponding address of the buffer 233. This buffer 233 has 8 rows and 8 columns of DCT coefficients D
It has a capacity equivalent to , and stores zero in advance as data corresponding to each component.

Therefore, as described above, the readout circuit 2
By storing each component read out in step 32 in the buffer 233, only the above-mentioned predetermined components are extracted from the DCT coefficients D corresponding to each block stored in the buffer 221, and the DC transform includes only these components as effective coefficients.
A T coefficient D can be obtained. The DCT coefficient D obtained in this way is an inverse DCT coefficient D corresponding to the inverse orthogonal transform means 113.
It is input to the CT conversion unit 241 and is converted to the inverse DCT conversion unit 24.
1, two-dimensional DCT transformation is performed. As a result, one block of image corresponding to the DCT coefficient D obtained by the above-described extraction means 112 is restored, and pixel data representing this image is sequentially input to the reduced image generation means 114.

The reduced image generating means 114 includes an extraction circuit 251 for extracting a part of input pixel data, a buffer 252 for holding the extracted pixel data, and an extraction operation by the extraction circuit 251 and a holding operation by the buffer 252. It is composed of an extraction control section 253 that controls operations. The extraction control unit 253 is configured to instruct the extraction circuit 251 to extract pixel data corresponding to the effective coefficients included in the DCT coefficients D obtained by the extraction means 112 described above. This extraction control unit 253 estimates the pixel (inflection point) at which the image data obtained by the inverse DCT transformation unit 241 described above takes an extreme value based on the basis vector corresponding to the effective coefficient described above, and the estimation result Pixels to be extracted may be determined based on this. For example, the above-mentioned extraction means 112 extracts the DC component and its surrounding three parts.
In the case of a configuration in which two AC components are extracted, it is estimated that there is no inflection point from the waveform of the first-order basis vector. In this case, it is sufficient to instruct the extraction circuit 251 to extract only the base point. In addition, by experiment, we determined the pixels where extreme values appear most frequently in the restored image data.
This pixel may be used as an inflection point.

The above-mentioned extraction control unit 253 also controls, for each block, the buffer 2 corresponding to the number of extracted pixels, depending on the position occupied by the block on the original image data.
52 areas are allocated and the addresses of these areas are sequentially designated as storage locations for the pixel data extracted by the extraction circuit 251. In this way, pixel data corresponding to the base point and inflection point are extracted from the image data of each block, a reduced image consisting of these pixels is generated, and as a collection of these reduced images, the image data of one screen is generated. A reduced image corresponding to is obtained. Here, the reduced image corresponding to each block does not need to be a square matrix; for example, if the numbers of base points and inflection points in the row direction and column direction are m and n, respectively, the reduced image consists of n×m pixels. All you have to do is generate an image.

The set of reduced images obtained in this way is input as one screen worth of image data to the transformation encoding means 115, which consists of a DCT transformation section 261, a linear quantization section 262, and an encoding section 263. , is variable-length coded in the same manner as the conventional ADCT method. For example, if a reduced image consisting of 4 pixels is generated corresponding to each block, DCT
By performing DCT transformation processing on each block of 8×8 pixels in the conversion unit 261, the reduced images corresponding to 16 blocks are DCT-transformed all at once and sent out as encoded data corresponding to one block. .

[0037] At this time, the encoding unit 263 may add information regarding the base point and inflection point constituting the reduced image to the beginning of one screen's worth of image data as encoded information, and then transmit the data. As mentioned above, by performing inverse DCT on the DCT coefficient D obtained by extracting only a part of the components from the DCT coefficient D, the image of the corresponding block can be obtained by superimposing the basis vectors corresponding to the extracted components. Approximate image data can be obtained. This removes high spatial frequency components and noise components contained in the original image,
The amount of information can be compressed, and by extracting only base points and inflection points from this image data, it is possible to obtain a set of reduced images that do not include redundant data that can be estimated from the waveform of the base vector. Can be done. By converting and encoding this set of reduced images, the amount of information is further compressed, so that the image data can be compressed and transmitted at a significantly higher compression rate than before.

[0038] This makes it possible to reduce the time required to transmit encoded data even when there is a restriction on the transmission speed of a line or the like, and to speed up the restoration process of image data. Furthermore, the DCT coefficients D input to the inverse DCT transform unit 241 described above include many invalid coefficients whose values are zero. Therefore, when performing matrix calculations corresponding to inverse DCT transformation processing, image data corresponding to each block is restored by executing only calculation processing related to effective coefficients and omitting calculation processing corresponding to invalid coefficients. The time required for processing can be shortened. For example, the inverse DCT transform unit 241 may be constructed using the technique disclosed in Japanese Patent Application No. 2-259484 entitled "Image Data Restoration Method and Image Data Restoration Apparatus" by the present applicant.

Furthermore, instead of providing a separate DCT conversion unit 261 that performs DCT transformation on the above-mentioned set of reduced images, the DCT conversion unit 211 converts the original image and the set of reduced images into DC
It may also be configured to perform T conversion. FIG. 6 is a block diagram of another embodiment of the image data compression device. As shown in FIG. 6, one screen worth of image data is input to the DCT converter 211 via the selector 271, and the conversion result by the DCT converter 211 is transferred to the buffer 222 via the demultiplexer 272.
The configuration is such that it is stored in Further, after the DCT conversion processing of the image data is completed by the DCT conversion unit 211, the selector 271 and the demultiplexer 272 described above are switched, and the reduced image stored in the buffer 252 is converted into DCT.
Input to the conversion unit 211 and DCT conversion unit 211
The configuration may be such that the output of is input to the linear quantization section 262. As a result, the DCT transform section 211 and the linear quantization section 2
62 and the encoding unit 263, the conversion encoding means 115
functions are realized.

In this case, since it is not necessary to provide two DCT conversion sections, the circuit scale of the image data compression apparatus does not increase. Note that a configuration may also be adopted in which the DCT coefficients D obtained by the DCT transformation unit 211 are once quantized and then inversely quantized, and a reduced image is generated based on the obtained DCT coefficients D.

FIG. 7 is a block diagram of an embodiment of the image data compression apparatus according to claim 8. In FIG. 7, the extraction means 112 of the image data compression apparatus has a configuration in which a linear quantization section 234 and a multiplier 235 are added to the extraction means 112 shown in FIG. This linear quantization section 234 corresponds to the quantization means 141, and includes a quantization threshold holding section 236 that holds quantization thresholds corresponding to each component of the DCT coefficient D.
Each component of the DCT coefficient D obtained by the DCT conversion unit 211 is divided by the corresponding quantization threshold value, and the result is stored in the buffer 221. Also,
In this case, the address calculation unit 231 instructs the read operation of the above-described predetermined components, and outputs the quantization threshold values corresponding to these components to the quantization threshold holding unit 2 described above.
36 may be used. Also, in response to this instruction, the quantization threshold output from the quantization threshold holding unit 236 is input to the multiplier 235, and the multiplier 235 multiplies the output of the readout circuit 232 by the quantization threshold described above. The structure is as follows. In this way, the quantization threshold holding section 236 and the multiplier 235 control the inverse quantization means 14.
Function 2 is realized, and by sequentially storing the dequantization results in the buffer 233, DCT coefficients D consisting of DCT coefficients obtained by dequantizing the quantized coefficients are obtained.

In this way, by once quantizing the DCT coefficients D and then inversely quantizing them, the noise components contained in the high spatial frequency components are particularly removed, and the DCT coefficients D containing no noise components are obtained. Obtainable. This results in
By eliminating noise components from the reduced image corresponding to each block, the amount of information can be reduced. Further, as shown in FIG. 8, the address calculation section 231 and the readout circuit 232 read predetermined DCT coefficients from the buffer 221 and input them to the linear quantization section 234.
The quantization result may be input to the multiplier 235 together with the corresponding quantization threshold.

In this case, only the predetermined components of the DCT coefficients D are selectively quantized, so it is possible to omit the division process for quantizing the components that are not extracted.
It is possible to reduce the time required for quantization processing and speed up image data compression processing. Incidentally, an image data compression device may receive an image whose gradation changes gradually as a whole, or an image whose gradation changes drastically. Furthermore, even in a single screen image, there may be a mixture of parts where the gradation changes are gradual and parts where the gradation changes are rapid. If we extracted the same number of DCT coefficients and generated a reduced image consisting of the same number of pixels for each block that makes up such an image, the amount of information would be excessive in areas where the gradation changes are gradual. Then,
On the other hand, there is a possibility that the amount of information will be insufficient in areas where gradation changes are severe.

[0044] A method of compressing image data while taking such image characteristics into account will be described below. Figure 9
FIG. 2 is a configuration diagram of an embodiment of an image data compression device according to claim 6; In FIG. 9, the image data compression apparatus is configured by adding a distribution determining section 281, a feature storage section 282, and an extraction instruction section 283 to the image data compression apparatus shown in FIG. The distribution discriminator 281 described above includes the linear quantizer 234
The configuration is such that the distribution pattern of the effective coefficients in the quantization coefficient DQU input from the block is determined, and the determination result is stored in the feature storage unit 282 as the feature of each block. Further, after the distribution pattern determination for all blocks is completed, the extraction instruction unit 28
3 divides a portion where blocks having the same or similar distribution pattern are consecutive into regions, and instructs the address calculation unit 231 described above to extract a component suitable for each region. . For example, extraction instruction section 2
83 may instruct extraction of components included in portions where effective coefficients are concentrated in the distribution pattern corresponding to each region.

In this case, the linear quantization section 234, the distribution discriminating section 281, the feature storage section 282, and the extraction instruction section 283 form the feature analysis means 121, and the distribution discriminating section 281
By inputting the quantization coefficient DQU into the
Image features can be analyzed without being affected by noise components included in the DCT coefficients D. Further, in response to an instruction from the above-mentioned extraction instruction section 283, the address calculation section 231 instructs the readout circuit 23 to perform the readout operation of the corresponding component.
You can instruct 2. In response to this, the extraction control unit 253 of the reduced image generation means 114 estimates the positions of the base point and the inflection point based on the above-mentioned components, and determines the pixels to be extracted. In addition, in response to the case where various DCT coefficient D components are extracted, estimated inflection points are estimated and held in advance, and the extraction control unit 253 The configuration may be such that the estimation result corresponding to the above is selected.

[0046] In this way, depending on the characteristics of each area,
It becomes possible to change the number of DCT coefficients and the number of pixels to be extracted. This makes it possible to flexibly handle cases where images with various characteristics are input, or cases where the input image contains parts with different characteristics. It is possible to prevent information from being sent.

Furthermore, in this case, the transform encoding means 115
The set of reduced images obtained as described above is variable-length encoded for each area, and information about the pixels extracted as the reduced image is placed at the beginning of the encoded data as encoded information for each area. Just add it and send it. Next, an image data decompression device that decompresses image data based on encoded data obtained by the above-described image data compression device will be described.

FIG. 10 is a block diagram of an embodiment of the image data restoration apparatus of the present invention. In FIG. 10, a decoding section 311, an inverse quantization section 312, and an inverse DCT transformation section 313 form a reduced image restoration means 151, and as in the conventional case,
Decode the input encoded data, inverse quantize it, and inverse DC
The configuration is such that a set of reduced images is restored by performing T transformation. The set of reduced images obtained by this reduced image restoration means 151 is configured to be sequentially stored in the buffer 321. In addition, in FIG. 10, the arithmetic processing section 331 and the arithmetic control section 332 form a block restoration means 152, which restores the pixel data of the corresponding block from the reduced image as described later. The configuration is such that each block is output as a restored image.

The above-mentioned decoding unit 311 restores the quantization coefficient DQU from the input encoded data in the same manner as in the past, and also extracts the encoded information added to each screen or each area. The encoded information is configured to be sent to the arithmetic control section 332 described above. Further, based on this encoded information, the calculation control unit 332 reads out the reduced image corresponding to each block from the buffer 321 and inputs it to the calculation processing unit 331.
The configuration is such that the arithmetic processing by 1 is controlled.

[0050] This arithmetic processing unit 331 is configured to obtain other pixel data of the corresponding block by complementing the change in pixel data between the base point and the inflection point constituting the input reduced image with a straight line. It becomes. In addition, the base vectors that contribute to the rough changes in the pixel data of each block are determined from the pixel data at the base point and the inflection point, and a function obtained by superimposing these base vectors is used to perform interpolation processing. It's okay.

[0051] In this way, a set of reduced images can be restored from the encoded data obtained by the above-described image data compression device, and image data of corresponding blocks can be restored from these reduced images. In addition, in the image data compression device, since reduced images corresponding to multiple blocks are transform-encoded at once, the decoding unit 311, inverse quantization unit 312, and inverse DCT transformation unit 313 By performing processing corresponding to a significantly smaller number of blocks, it is possible to obtain a set of reduced images corresponding to each block constituting one screen. That is,
Decoding section 311, inverse quantization section 312, and inverse DCT transformation section 31
Since the amount of arithmetic processing in step 3 is significantly reduced, the time required for restoration processing can be significantly reduced. For example, when a reduced image consisting of 2×2 pixels is generated by the above-mentioned reduced image generation means 114, the reduced images corresponding to 16 blocks are converted and encoded at once. Therefore, the decoding section 311, the inverse quantization section 312, and the inverse DC
The number of blocks to be processed by the T transform unit 313 is
Since this is 1/16th of the conventional amount, the amount of calculation by each of these parts can be reduced to 1/16th by simple calculation. Further, the calculation processing by the calculation processing unit 331 of the block restoration means 152 is simpler calculation processing and has a smaller amount of calculation than the inverse DCT transformation processing, whereas the amount of calculation reduced as described above is Very large. Therefore, since the arithmetic processing unit 331 is added, the amount of computation does not increase as a whole compared to the conventional image data restoring device, and it is possible to speed up the image data restoring process.

[0052] In this way, when a set of reduced images is transform encoded, the image data is compressed at a much higher compression rate than before, reducing the time required to transmit the encoded data, and also reducing the time required for the restoration process. It is possible to reduce the time required for the restoration process by reducing the calculation processing required for the restoration process. By the way, transform encoding methods such as the ADCT method described above are non-information preserving encoding methods, and in the encoding process, part of the information regarding the set of reduced images is lost, resulting in significant deterioration of image quality. I end up. For this reason, image data compression devices and image data decompression devices that use the above-mentioned transform encoding method cannot perform compression that is higher than image quality, such as when performing hierarchical restoration using a transfer line with limited transmission speed. Suitable for applications where efficiency is a priority.

On the other hand, there are also applications that require a high compression rate and do not allow deterioration in image quality. In the following, a method for obtaining a high compression rate while maintaining image quality in order to support such uses will be described. FIG. 11 is a block diagram of an embodiment of an image data compression apparatus according to a seventh aspect of the present invention. In FIG. 11, the image data compression apparatus includes predictive encoding means 131 in place of the transform encoding means 115 of the image data compression apparatus shown in FIGS. 5 to 9. This predictive encoding means 131 includes, for example, a difference circuit 291 that sequentially calculates previous value differences for pixel data stored in the buffer 252 of the reduced image generation means 114, and a code that performs variable length encoding on the obtained difference data. The configuration may include a conversion section 292.

In this case, the predictive encoding means 13 described above
1, the difference data of the reduced image is encoded as is, so information regarding the reduced image is encoded without being lost. Therefore, by encoding the reduced image using the predictive encoding method, which is an information-preserving encoding method, it becomes possible to accurately restore the reduced image on the restoration side, and the deterioration of image quality is reduced. can be suppressed.

[0055] Also in this case, since the encoded data does not include redundant data that can be estimated from the base point and inflection point that make up the reduced image, the conventional ADC
A higher compression ratio than the T method can be obtained. The encoded data obtained in this way can be restored by the image data restoration device shown in FIG.

In FIG. 12, the image data restoration device is
In place of the inverse quantization section 312 and the inverse DCT transformation section 313 shown in FIG. 10, the reduced image restoration means 151 includes an integration circuit 314, and the output of the integration circuit 314 is sequentially stored in the buffer 321. And it is sufficient. In this case, since the above-mentioned difference data itself is obtained as the decoding result by the decoding unit 311, the reduced image can be restored by sequentially integrating the difference data by the integration circuit 314. Further, the reduced images obtained in this manner are each restored into corresponding blocks by the above-mentioned block restoration means 152 in the same manner as in the case where the transform encoding method is applied.

The integration process by the integration circuit 314 is a simple calculation process, and the amount of calculation is very small compared to the inverse quantization process and the inverse DCT transformation process. Compared to the case where this method is used, the time required for restoration processing of a reduced image can be shortened. This makes it possible to speed up image data restoration processing.

[0058]

[Effects of the Invention] As explained above, the present invention provides a method for reducing pixels that can be approximately determined by re-encoding a set of reduced images generated based on an image restored from an extraction coefficient matrix as image data. It becomes possible to obtain encoded data that does not include redundant data related to the image data, thereby shortening the time required to transmit the encoded data and speeding up image data restoration processing.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing the principle of the invention.

FIG. 2 is a configuration diagram of the invention according to claim 3.

FIG. 3 is a configuration diagram of the invention according to claim 7.

FIG. 4 is a configuration diagram of the invention according to claim 9.

FIG. 5 is a configuration diagram of an embodiment of an image data compression device of the present invention.

FIG. 6 is a configuration diagram of another embodiment of the image data compression device of the present invention.

FIG. 7 is a configuration diagram of an embodiment of an image data compression device according to an eighth aspect of the invention.

FIG. 8 is a configuration diagram of another embodiment of the image data restoration device according to the invention of claim 8;

FIG. 9 is a configuration diagram of an embodiment of an image data compression device according to a sixth aspect of the invention.

FIG. 10 is a configuration diagram of an embodiment of an image data restoration device of the present invention.

FIG. 11 is a configuration diagram of an embodiment of an image data compression device according to a seventh aspect of the invention.

FIG. 12 is a configuration diagram of another embodiment of the image data restoration device of the present invention.

FIG. 13 is a configuration diagram of a conventional image data compression device.

FIG. 14 is a diagram showing an example of blocks.

FIG. 15 is a diagram showing an example of DCT coefficients D. FIG.

FIG. 16 is a diagram showing a quantization matrix.

FIG. 17 is a diagram showing a quantization coefficient DQU.

FIG. 18 is an explanatory diagram of a zigzag scan.

FIG. 19 is a configuration diagram of a conventional image data restoration device.

FIG. 20 is an explanatory diagram of basis vectors.

[Explanation of symbols]

111 Orthogonal transformation means 112 Extraction means 113 Inverse orthogonal transformation means 114 Reduced image generation means 115 Transformation encoding means 121 Feature analysis means 131 Predictive encoding means 141 Quantization means 142 Inverse quantization means 151 Reduced image restoration means 152 Block restoration means 211 , 261, 611 DCT conversion unit 221, 23
3, 252, 321 Buffer 231 Address calculation section 232 Read circuit 234, 262, 620 Linear quantization section 235 Multiplier 236, 623 Quantization threshold holding section 241, 313, 731 Inverse DCT transformation section 251
Extraction circuit 253 Extraction control section 263, 292, 631 Encoding section 271 Selector 272 Demultiplexer 281 Distribution discriminator 282 Feature storage section 283 Extraction instruction section 291 Difference circuit 311, 711 Decoding section 312, 720 Dequantization section 314 Integration circuit 331 Arithmetic processing unit 332 Arithmetic control unit 621 DCT coefficient input unit 622 Divider 632 Code table 712 Decoding table

Claims (9)

[Claims] 1. Image data representing a multivalued image is orthogonally transformed for each block of M×N pixels to obtain a coefficient matrix of M rows and N columns each consisting of transform coefficients, and the coefficient matrix corresponding to each block is A plurality of components are extracted from each block, an M-by-N extraction coefficient matrix is generated that includes only the extracted components as effective coefficients having values other than zero, and the extraction coefficient matrix is inversely orthogonally transformed. restore the image of
A plurality of pixels are each extracted from the obtained image, a reduced image consisting of these pixels is generated corresponding to each block, and a set of reduced images corresponding to each block is re-encoded as image data. An image data compression method characterized by compression. 2. Image data representing a multivalued image is orthogonally transformed for each block to obtain the coefficient matrix, and the characteristics of the multivalued image are analyzed based on the coefficient matrix corresponding to each block. Extracting components corresponding to the features of the multivalued image from the coefficient matrix corresponding to each block to generate the extraction coefficient matrix, and performing inverse orthogonal transformation on the extraction coefficient matrix to restore the image of each block; Pixels corresponding to the characteristics of the multivalued image are extracted from the obtained image, a reduced image consisting of these pixels is generated corresponding to each block, and a set of reduced images corresponding to each block is created as an image. An image data compression method characterized by re-encoding and compressing the data. 3. Orthogonal transformation means (111) for orthogonally transforming image data representing a multivalued image for each block to obtain the coefficient matrix, and extracting a plurality of components from the coefficient matrix corresponding to each block. an extraction means (112) for generating the extraction coefficient matrix corresponding to each block; an inverse orthogonal transformation means (113) for restoring an image of each block by inversely orthogonally transforming each of the extraction coefficient matrices; Pixels corresponding to the plurality of components extracted by the extraction means (112) are extracted from the image data obtained by the inverse orthogonal transformation means (113), and a reduced image consisting of these pixels is created corresponding to each block. An image data compression device comprising: reduced image generation means (114) for generating reduced images; and conversion encoding means (115) for converting and encoding image data consisting of reduced images corresponding to each of the blocks. 4. The image data compression apparatus according to claim 3, wherein the extraction means (112) extracts a plurality of components including a DC component and surrounding AC components of the coefficient matrix to generate an extracted coefficient matrix. An image data compression device characterized by being configured to generate image data. 5. The image data compression apparatus according to claim 3, wherein the reduced image generation means (114) performs the following steps based on base vectors corresponding to a plurality of components included as effective coefficients in the extraction coefficient matrix: An image data compression device characterized by having a configuration that determines pixels to be extracted. 6. The image data compression apparatus according to claim 3, further comprising a feature analysis means (121) for analyzing features of the multivalued image based on a coefficient matrix corresponding to each block.
), wherein the extraction means (112) is configured to determine a component to be extracted from the coefficient matrix according to an analysis result by the feature analysis means (121). 7. The image data compression apparatus according to claim 3, wherein the transform encoding means (115) is replaced by a predictive encoding means (115) for predictively encoding image data consisting of reduced images corresponding to each block. 131). 8. The image data compression apparatus according to claim 3, wherein the extraction means (112) performs quantization to obtain quantized coefficients by quantizing the input coefficient matrix components using a corresponding quantization threshold. means (141), and dequantization means (142) for dequantizing input quantization coefficients using a corresponding quantization threshold, and comprising transform coefficients obtained by dequantizing the quantization coefficients. An image data compression device characterized by being configured to generate an extraction coefficient matrix. [Claim 9] Based on input encoded data,
reduced image restoring means (15) for restoring the set of reduced images;
1), and based on the pixel data of a plurality of pixels forming each reduced image obtained by the reduced image restoring means (151), find the pixel data of other pixels forming the corresponding block, and An image data restoring device comprising: block restoring means (152) for restoring blocks.