JP7809337B2 - Encoder, data processing method, data processing program, data processing system, and decoder - Google Patents

Description

The present invention relates to encoding technology for data to be processed, such as audio data and image data.

Conventionally, a hybrid compression technique has been known for compressing image and audio data. This technique uses information obtained through mathematical modeling such as prediction or frequency decomposition as the symbols to be encoded, and assigns the shortest possible codes to the symbols using entropy coding (minimum redundancy coding).

Markov model compression is also known as one of the key technologies for compressing image and audio data. Markov model compression is a modeling technique that uses the correlation between preceding and succeeding symbols in an input sequence. In other words, it defines the probability of a symbol appearing after a specific symbol sequence, and the number of elements in the preceding symbol sequence is used as the order of the model. For this reason, first-order Markov model compression has an occurrence probability distribution for each symbol, and is implemented using multiple encoders and decoders.

For example, Non-Patent Document 1 discloses that text data and image data can be encoded with a small amount of code using first-order Markov model compression that uses canonical Huffman coding and arithmetic coding as entropy coding.

Other key technologies for compressing image and audio data include hierarchical subband decomposition and hierarchical subband coding, an encoding method that uses it. Subband coding is a method in which a signal is divided into several subbands, each of which is then quantized and entropy coded. Subband decomposition is a type of modeling that uses frequency decomposition. In particular, when dividing image or audio data into subbands, hierarchical subband decomposition is sometimes used, in which subband decomposition is recursively repeated on the low-frequency band signal obtained by the division, due to reasons such as the high energy in the low-frequency band.

For example, Patent Document 1 discloses a technology that enables image compression and decompression with a small amount of calculations based on hierarchical subband coding that utilizes hierarchical subband division. In Patent Document 1, the Hadamard transform is used for the subband division.

Furthermore, Non-Patent Document 2 discloses that JPEG2000 employs hierarchical subband decomposition (discrete wavelet transform) for frequency transformation.

Japanese Patent Application Laid-Open No. 2004-289290

Masayuki Tanemura, Ryuji Tokunaga, "Improving the Coding Efficiency of First-Order Markov Compression Based on Semi-Static Codes," Transactions of the Institute of Electronics, Information and Communication Engineers (A), vol. J-88-A, No. 12, pp. 1552-1559 (2005) C. Christopoulos, A. Skodras, and T. Ebrahimi. The jpeg2000 still image coding system: an overview. IEEE Transactions on Consumer Electronics, Vol.46, No.4, pp.1103-1127,2000

When semi-static codes are used in the encoding process, additional information is required to describe the codebook. In particular, Markov model compression using semi-static codes requires multiple encoders and decoders, which creates the problem of an enormous amount of additional information.

In addition, when performing hierarchical subband coding, spatiotemporal sequence data with different statistical properties is generated for each band. If a single Markov model compressor is used to encode data sequences with widely differing statistical properties as symbols, appropriate code assignment cannot be performed, resulting in a deterioration in coding efficiency. On the other hand, if a different Markov model compressor is used for each band, it is necessary to provide an encoder and decoder for each band, which leads to an even greater increase in additional information and a deterioration in coding efficiency.

The present invention was developed in consideration of the above circumstances, and its purpose is to provide technology that can improve the compression rate of data to be compressed, such as image data and audio data.

To achieve the above object, one aspect of the encoder is an encoder that compresses specified data to be processed, and includes: a frequency band division unit that divides the data to be processed into multiple partial frequency bands; a category division unit that, for at least some of the partial frequency bands, divides multiple encoding target symbols included in elements to be compressed based on the elements of the partial frequency bands into multiple categories based on the energy of the symbol immediately preceding the encoding target symbol; and an encoding unit having multiple encoders that encode the encoding target symbols included in the elements to be compressed based on the elements of the multiple partial frequency bands, at least one of which is a common encoder that encodes the encoding target symbols of elements to be compressed included in one of the categories and the encoding target symbols of elements to be compressed in a partial frequency band that is strongly associated with that category using a common code.

This invention makes it possible to improve the compression rate of data to be compressed.

FIG. 1 is a diagram illustrating an outline of subband division according to an embodiment. FIG. 2 is a diagram illustrating the first encoding process according to the embodiment. FIG. 3 is a diagram illustrating the second encoding process according to the outline of the embodiment. FIG. 4 is a diagram showing the overall configuration of an image processing system according to an embodiment. FIG. 5 is a diagram illustrating the configuration of a subband division unit according to an embodiment. FIG. 6 is a diagram illustrating the configuration of a partial subband division unit according to an embodiment. FIG. 7 is a configuration diagram of a prediction residual processing unit according to an embodiment. FIG. 8 is a configuration diagram of a partial prediction residual processing unit of one layer according to an embodiment. FIG. 9 is a configuration diagram of a partial prediction residual processing unit of another layer according to an embodiment. FIG. 10 is a diagram illustrating an example of division into bands and categories according to an embodiment. FIG. 11 is a diagram illustrating another example of division into bands and categories according to an embodiment. FIG. 12 is a diagram illustrating a first configuration example of an encoding unit according to an embodiment. FIG. 13 is a diagram illustrating a second configuration example of the encoding unit according to an embodiment. FIG. 14 is a diagram illustrating a first configuration example of a decoding unit according to an embodiment. FIG. 15 is a diagram illustrating a second configuration example of a decoding unit according to an embodiment. FIG. 16 is a block diagram of a reproducing unit according to an embodiment. FIG. 17 is a diagram illustrating the configuration of one partial reproduction unit according to one embodiment. FIG. 18 is a configuration diagram of another partial reproducing unit according to an embodiment. FIG. 19 is a flowchart of an encoding process according to an embodiment. FIG. 20 is a flowchart of a decoding process according to one embodiment. FIG. 21 is a diagram illustrating the configuration of a computer device according to an embodiment.

Embodiments will be described with reference to the drawings. Note that the embodiments described below do not limit the invention as defined in the claims, and not all of the elements and combinations thereof described in the embodiments are necessarily essential to the solution of the invention.

First, we will explain the overview of the embodiment. Figure 1 is a diagram explaining the overview of subband division according to the embodiment.

In this embodiment, the encoder that performs entropy encoding is shared across data in different frequency bands. This allows for appropriate code assignment while minimizing the amount of additional information required for decoding, thereby improving encoding efficiency (compression efficiency).

Natural image and audio data (signals) have the following properties: (1) Small signals tend to exist when they are located spatiotemporally close to signals with small energy. (2) Large signals tend to exist when they are located spatiotemporally close to signals with large energy. (3) High-frequency bands have less energy than low-frequency bands.

Taking this property into consideration, (1) a common encoder is used to encode signals that are located spatiotemporally close to small signals in the low-frequency band, and (2) a common encoder is used to encode signals in the high-frequency band or signals that are located spatiotemporally close to large signals in the high-frequency band.

The data processing system 200 according to this embodiment includes a target data input unit 201 that inputs data to be compressed (target data), and a subband decomposition unit 202 that performs hierarchical subband decomposition on the target data to divide it into data of multiple bands.

The subband division unit 202 includes multiple filters 211 that divide the data into bands, and a thinning unit 212 that adjusts the sampling interval in the bands divided by the filters 211 to thin out the data. Here, as shown in Figure 1, the lower frequency band of two adjacent bands is designated as band A, and the higher frequency band is designated as band B, and the following encoding process is described.

In the encoding process, bands A and B are mathematically modeled and quantized to determine the symbol sequence to be encoded that describes the signal. Once the symbol sequence to be encoded has been determined, the symbols in band A are divided into at least two categories. Here, L refers to the symbol most likely to produce the next smallest signal, in other words, the category containing one or more symbols to be encoded whose previous symbol has small energy, and H refers to the symbol most likely to produce the next largest signal, in other words, the category containing one or more symbols to be encoded whose previous symbol has large energy.

The data processing system 200 then creates a codebook by counting the number of times each encoding target symbol appears after a symbol belonging to each category appears.

Figure 2 is a diagram illustrating the first encoding process according to an embodiment.

Codebook 220 includes columns for band 220a, category 220b, symbol 220c, and occurrence count 220d. Band 220a stores the band name of each divided subband. Category 220b stores the category divided in each subband band. Symbol 220c stores each encoding target symbol that may appear after a symbol belonging to the category. Occurrence count 220d stores the number of times that the encoding target symbol of symbol 220c appears after a symbol of category 220b in a row (occurrence count). Based on this codebook 220, it is possible to determine the codeword to encode each encoding target symbol. Note that the technology for creating codewords from the occurrence count is well known, so a description thereof will be omitted.

Here, the data processing system 200 assigns an encoder that performs entropy encoding to each band category, and uses each encoder to encode the symbol string of that category based on the codeword. Next, the data processing system 200 creates compressed data with additional information added to identify the codeword for the encoded data encoded by the encoder, and transmits this data to the decoder. Note that, as shown in Figure 2, the additional information is the number of occurrences when arithmetic coding is used, or, for example, the code length of each symbol when canonical Huffman coding is used.

The data processing system 200 may also create the codebook shown in Figure 3.

Figure 3 is a diagram illustrating the second encoding process according to an embodiment.

Codebook 230 includes columns for band 230a, category 230b, symbol 230c, and number of occurrences 230d. Each column is the same as the column with the same name in codebook 220. In codebook 230, the number of times each symbol to be coded appears after a symbol in the L category in band A is added together with the number of times each symbol to be coded appears after a symbol in the H category in adjacent band B, which has a strong correlation with band A.

In this case, the data processing system 200 encodes multiple categories (symbols in the H category of band B and the L category of band A) based on codewords using a single encoder (common encoder). For categories that are not combined, an encoder is assigned to each band category for encoding. Next, the data processing system 200 creates compressed data with additional information added to identify the codewords for the encoded data encoded by the encoder, and transmits the compressed data to the decoder. Note that, as shown in FIG. 3, the additional information is the number of occurrences when arithmetic coding is used, or the code length of each symbol when canonical Huffman coding is used, for example. This second encoding process can reduce the number of encoders required and the amount of additional information data, thereby improving the compression rate of the compressed data transmitted to the decoder.

In addition, the decoder side of the data processing system 200 can create a codebook using the additional information, which can identify the codeword assignment used by the encoder. The decoder decodes a symbol string from the code string of the encoded data of the compressed data based on the identified codeword assignment rules.

Next, we will explain in detail the image processing system according to one embodiment.

Figure 4 is a diagram showing the overall configuration of an image processing system according to one embodiment.

Image processing system 1 is an example of a data processing system and includes an encoder 10 and a decoder 30. Encoder 10 and decoder 30 are connected via a network 50. Network 50 is, for example, a local area network (LAN) or a wide area network (WAN).

The encoder 10 includes a target data input unit 11, a subband decomposition unit 12 as an example of a frequency band decomposition unit, a prediction residual processing unit 14, an encoding control unit 15 as an example of a category decomposition unit and determination unit, an encoding unit 16, and a compressed data output unit 18 as an example of a data transmission unit.

The target data input unit 11 inputs an image to be compressed (data to be processed). The target image may be, for example, a still image or a single frame of a moving image. The target image may be obtained from the auxiliary storage device 106 (see Figure 21) within the encoder 10, or may be obtained from an imaging device (not shown).

The subband decomposition unit 12 performs hierarchical subband decomposition on the data to be processed, dividing it into data of multiple bands.

Figure 5 is a diagram showing the configuration of a subband division unit according to one embodiment.

The subband decomposition unit 12 includes multiple partial subband decomposition units 12-n (12-0, 12-1, 12-2, 12-3).

The partial subband division unit 12-n receives input data δ _n and outputs α _n+1 , β _n+1 , γ _n+1 , and δ _n+1 .

Here, α _n+1 is horizontal high frequency band data for the input data δ _n , β _n+1 is vertical high frequency band data for the input data δ _n , γ _n+1 is diagonal high frequency band data for the input data δ _n , δ _n+1 is low frequency band data for the input data δ _n , and δ ₀ is the processing target data input by the target data input unit 11.

Figure 6 is a diagram showing the configuration of a partial subband division unit according to one embodiment.

The partial subband decomposition unit 12-n includes a vertical low-pass filter (h vertical) 121, a vertical thinning unit (2↓ vertical) 122, a horizontal low-pass filter (h horizontal) 123, a horizontal thinning unit (2↓ horizontal) 124, a horizontal high-pass filter (g horizontal) 125, a horizontal thinning unit 126, a vertical high-pass filter (g vertical) 129, a vertical thinning unit 130, a horizontal low-pass filter 131, a horizontal thinning unit 132, a horizontal high-pass filter 133, and a horizontal thinning unit 134.

The vertical low-pass filter 121 performs a process of applying a low-pass filter to the input data δ _n in the vertical direction. Specifically, the vertical low-pass filter 121 performs a convolution process on the input data δ _n using a function h[x] shown in equation (1).

Here, if the value (data value) of the input data _δn at the vertical coordinate m is S[m], the result (s*h)[m] obtained by the convolution process at the coordinate m is expressed as shown in equation (2).

In this convolution process, a value related to the sum of the data values of adjacent coordinates is calculated for the input data. The result of this calculation corresponds to the low-frequency components in the input data.

The vertical high-pass filter 129 performs a process of applying a high-pass filter to the input data δ _n in the vertical direction. Specifically, the vertical high-pass filter 129 performs a convolution process on the input data δ _n using a function g[x] shown in equation (3). Here, when a Haar wavelet basis is used as the functions h and g, this is called a Hadamard transform.

Here, if the data value for the vertical coordinate m of the input data _δn is S[m], the result (s*g)[m] obtained by the convolution process at the coordinate m is expressed by equation (4).

In this convolution process, a value related to the difference between data values at adjacent coordinates is calculated for the input data. The result of this calculation corresponds to the high-frequency components in the input data.

The vertical thinning units 122 and 130 perform a 2:1 thinning process on the input data in the vertical direction, i.e., thinning two data values in the vertical direction to one data value.

The horizontal low-pass filters 123 and 131 apply a low-pass filter to the input data in the horizontal direction. The processing by the horizontal low-pass filters 123 and 131 is the same as the vertical processing by the vertical low-pass filter 121, but is instead horizontal processing.

The horizontal high-pass filters 125 and 133 apply a high-pass filter to the input data in the horizontal direction. The processing by the horizontal high-pass filters 125 and 133 is the same as the vertical processing by the vertical high-pass filter 129, but is instead horizontal processing.

The horizontal thinning units 124, 126, 132, and 134 perform a 2:1 thinning process on the input data in the horizontal direction, i.e., thinning two data values in the horizontal direction to one data value.

In the partial subband division unit 12-n, input data δ _n is converted to low-frequency band data δ _n+1 via a vertical low-pass filter 121, a vertical thinning unit 122, a horizontal low-pass filter 123, and a horizontal thinning unit 124, and then output. Also, in the partial subband division unit 12-n, input data δ _n is converted to horizontal high-frequency band data α _n +1 via a vertical low-pass filter 121, a vertical thinning unit 122, a horizontal high-pass filter 125, and a horizontal thinning unit 126, and then output. Also, in the partial subband division unit 12-n, input data δ _n is converted to vertical high-frequency band data β _n+1 via a vertical high-pass filter 129, a vertical thinning unit 130, a horizontal low-pass filter 131, and a horizontal thinning unit 132, and then output. Furthermore, in partial subband decomposition section 12-n, input data δ _n is converted into and output from diagonal high-frequency band data γ _n+1 via vertical high-pass filter 129, vertical thinning section 130, horizontal high-pass filter 133, and horizontal thinning section 134. For example, if partial subband decomposition section 12-n converts input data δ _n into 640×640 two-dimensional data, horizontal high-frequency band data α _n+1 , vertical high-frequency band data β _n+1 , and diagonal high-frequency band data γ _n+1 become 320×320 two-dimensional data.

Returning to the explanation of Figure 4, the prediction residual processing unit 14 performs prediction residual processing to calculate the difference between the predicted value and the data for each band (prediction residual: element to be compressed).

Figure 7 is a diagram showing the configuration of a prediction residual processing unit according to one embodiment.

The prediction residual processing unit 14 includes a partial prediction residual processing unit 14-4, a partial prediction residual processing unit 14-3, a partial prediction residual processing unit 14-2, and a partial prediction residual processing unit 14-1.

The partial prediction residual processing unit 14-4 receives {α ₄ , β ₄ , γ ₄ , δ ₄ } as input, and outputs {α ^· ₄ , β ^· ₄ , γ ^· ₄ }, δ ^· ₄ , and δ' _3. The partial prediction residual processing unit 14-3 receives {α ₃ , β ₃ , γ ₃ } and δ' ₃ as input, and outputs {α ^· ₃ , β ^· ₃ , γ ^· ₃ } and δ' _2. The partial prediction residual processing unit 14-2 receives {α ₂ , β ₂ , γ ₂ } and δ' ₂ as input, and outputs {α ^· ₂ , β ^· ₂ , γ ^· ₂ } and δ' ₁ . The partial prediction residual processing unit 14-1 receives {α ₁ , β ₁ , γ ₁ } and δ' ₁ as input, and outputs {α ^· ₁ , β ^· ₁ , γ ^· ₁ }.

Here, {α ^· _n , β ^· _n , γ ^· _n , δ ^· _n } is the quantized prediction residual for {α _n , β _n , γ _n , δ _n }, and δ' _n is the reconstructed value of δ _n at the decoder.

The prediction residual processing unit 14 outputs {α ^· ₁ , β ^· ₁ , γ ^· ₁ }, {α ^· ₂ , β ^· ₂ , γ ^· ₂ }, {α ^· ₃ , β ^· ₃ , γ ^· ₃ }, {α ^· ₄ , β ^· ₄ , γ ^· ₄ }, and δ ^· ₄ .

Figure 8 is a diagram showing the configuration of a partial prediction residual processing unit for one layer according to one embodiment.

The partial prediction residual processing unit 14-4 includes a DPCM (differential pulse code modulation) encoder 141, an ACP (alternating current component prediction) unit 142, a calculation unit 143, a quantization unit 144, an inverse quantization unit 145, a calculation unit 146, an ACP unit 147, and a partial subband synthesis unit 148.

The DPCM encoder 141 is capable of performing differential pulse code modulation processing, inputting δ ₄ and outputting a reproduced value δ' ₄ at the decoder and a prediction residual δ ^· ₄ .

The ACP units 142 and 147 receive the reproduction value δ' ₄ and calculate {α ^∼ ₄ , β ^∼ ₄ , γ ^∼ ₄ } by AC component prediction. {α ^∼ ₄ , β ^∼ ₄ , γ ^∼ ₄ } are predicted values for {α ₄ , β ₄ , γ ₄ } predicted by AC component prediction. The AC component prediction disclosed in Patent Document 1 can be used as the AC component prediction. The ACP units 142 and 147 may also receive δ ₄ and calculate {α ^∼ ₄ , β ^∼ ₄ , γ ^∼ ₄ } by AC component prediction from this δ ₄ .

The calculation unit 143 calculates the prediction residual by subtracting the predicted values {α ⁴ , β ₄ , γ ₄ } from {α ₄ , ^β ₄ , _γ ⁴ _} .

The quantization unit 144 quantizes the prediction residuals calculated by the calculation unit 143 and outputs {α ^· ₄ , β ^· ₄ , γ ^· ₄ }. The inverse quantization unit 145 inverse quantizes {α ^· ₄ , β ^· ₄ , γ ^· ₄ } and outputs the result.

The calculation unit 146 adds the predicted values {α ^∼ ₄ , β ^∼ ₄ , γ ^∼ ₄ } to the result of the inverse quantization by the inverse quantization unit 145 to calculate the reproduced values {α' ₄ , β' ₄ , γ' ₄ }.

The partial subband synthesis unit 148 receives the reproduction value δ' ₄ and the reproduction values {α' ₄ , β' ₄ , γ' ₄ }, performs the inverse transformation of the subband division process in the partial subband division unit 12-n, and outputs the reproduction value δ' ₃ .

Figure 9 is a diagram showing the configuration of a partial prediction residual processing unit for another layer according to one embodiment.

The partial prediction residual processing unit 14-n (here, n is an integer from 1 to 3) includes an ACP (alternating current component prediction) unit 142, a calculation unit 143, a quantization unit 144, an inverse quantization unit 145, a calculation unit 146, an ACP unit 147, and a partial subband synthesis unit 148. Note that the partial prediction residual processing unit 14-n does not include a DPCM encoder 141 because it can obtain the reproduction value δ' _n at the decoder from the lower-order partial prediction residual processing unit 14-n+1.

The ACP units 142 and 147 receive the reproduction value δ' _n and calculate {α ^∼ _n , β ^∼ _n , γ ^∼ _n } by AC component prediction. {α ^∼ _n , β ^∼ _n , γ ^∼ _n } are predicted values for {α _n , β _n , γ _n } predicted by AC component prediction. The AC component prediction disclosed in Patent Document 1 can be used as the AC component prediction.

The calculation unit 143 calculates the prediction residual by subtracting the predicted values {α ^∼ _n , β ^∼ _n , γ ^∼ _n } from {α _n , β _n , γ _n }.

The quantization unit 144 quantizes the prediction residuals calculated by the calculation unit 143 and outputs {α ^· _n , β ^· _n , γ ^· _n }. The inverse quantization unit 145 inverse quantizes {α ^· _n , β ^· _n , γ ^· _n } and outputs the result.

The calculation unit 146 adds the predicted values {α ^∼ _n , β ^∼ _n , γ ^∼ _n } to the result of the inverse quantization by the inverse quantization unit 145 to calculate {α' _n , β' _n , γ' _n }.

The partial subband synthesis unit 148 receives the reproduction value δ' _n and the reproduction values {α' _n , β' _n , γ' _n }, performs the inverse transformation of the subband division process in the partial subband division unit 12-n, and outputs a reproduction value δ' _n-1 .

The encoding control unit 15 controls the processing in the encoding unit 16, thereby controlling the processing of encoding the quantized prediction residuals {α ^· ₁ , β ^· ₁ , γ ^· ₁ }, {α ^· ₂ , β ^· ₂ , γ ^· ₂ }, {α ^· ₃ , β ^· ₃ , γ ^· ₃ }, {α ^· ₄ , β ^· ₄ , γ ^· ₄ }, and δ ^· ₄ output by the prediction residual processing unit 14. Here, the encoding control unit 15 may control the encoding unit 16 to encode δ ^· ₄ using a known technique, for example, the first-order Markov compression of Non-Patent Document 1. In addition, in this embodiment, the encoding control unit 15 controls the encoding unit 16 _{to use} the values ( _absolute values) excluding the positive and negative signs as the symbols to be encoded for ^{{ α ·} ₁ , β ^· ₁ ^, γ ^· ₁ }, {α ^· ₂ , β ^· ₂ , γ ^· ₂ }, {α ^· ₃ , β ^· ₃ , γ ^{·3}, and {α·4, β·4, γ·} ₄ }, and to encode the information indicating each positive and negative sign using known technology.

The encoding control unit 15 treats sub-blocks of different hierarchical levels, {α ^· ₁ , β ^· ₁ , γ ^· ₁ }, {α ^· ₂ , β ^· ₂ , γ ^· ₂ }, {α ^· ₃ , β ^· ₃ , γ ^· ₃ }, and {α ^· ₄ , β ^· ₄ , γ ^· ₄ }, as separate partial frequency bands. In this embodiment, {α ^· ₁ , β ^· ₁ , γ ^· ₁ } is band 1, {α ^· ₂ , β ^· ₂ , γ ^· ₂ } is band 2, {α ^· ₃ , β ^· ₃ , γ ^· ₃ } is band 3, and {α ^· ₄ , β ^· ₄ , γ ^· ₄ } is band 4.

The encoding control unit 15 determines the encoding target symbol sequence that describes each data. Because there is little information in the high-frequency band, the encoding control unit 15 does not divide band 1 into categories based on the energy of the symbol immediately preceding the encoding target symbol. Furthermore, for bands 2 to 4, the encoding control unit 15 divides bands 2 to 4 into multiple categories based on the energy of the symbol immediately preceding the encoding target symbol (e.g., the absolute value of the amplitude corresponding to the symbol). In this embodiment, for example, if the energy value of the immediately preceding symbol is equal to or less than a predetermined threshold θ, the symbol is classified as category L, and if the energy value of the immediately preceding symbol is greater than the threshold θ, the symbol is classified as category H. The threshold θ used to divide the categories may differ for each band.

The encoding control unit 15 then counts the number of times a symbol belonging to each category appears immediately before the symbol to be encoded and creates the codebook 150.

Figure 10 is a diagram illustrating an example of band and category division according to one embodiment.

Codebook 150 includes columns for band 150a, category 150b, symbol 150c, and occurrence count 150d. Band 150a stores the band name of each divided subband. Category 150b stores the category divided in each subband band. Symbol 150c stores each encoding target symbol that may appear after a symbol belonging to the category. Occurrence count 150d stores the number of times (occurrence count N) that the encoding target symbol of symbol 150c appears after a symbol of category 150b in a row. Note that the subscript of the occurrence count N of category 150b indicates the type of category (L or H), the first parameter in parentheses indicates the band number, and the second parameter indicates the symbol. Based on this codebook 150, a codeword to encode each encoding target symbol can be determined.

The encoding control unit 15 assigns seven encoders to the encoding unit 16 so that one encoder is assigned to band 1, which is not divided into categories, and one encoder is assigned to each of the categories in bands 2 to 4 (first configuration), and causes the encoding unit 16 in the first configuration to encode the target symbol sequence based on a codeword. The encoding control unit 15 also specifies the size of the compressed data (first compressed data) that includes the encoded data when encoding in the first configuration is performed and additional information for identifying the codeword.

In addition, the encoding control unit 15 creates a codebook 151 that assumes the case where one band or its category is grouped together with a closely related band (an adjacent band, or a band in an adjacent layer) or its category, and a common encoder is used.

Figure 11 is a diagram illustrating another example of band and category division according to one embodiment.

Codebook 151 includes columns for band 151a, category 151b, symbol 151c, and number of occurrences 151d. Each column is the same as the column with the same name in codebook 150. In codebook 151, the number of times each symbol to be coded in band 1 appears is added together with the number of times the same symbol to be coded in the L category of band 2; the number of times each symbol to be coded in the H category of band 2 appears together with the number of times the same symbol to be coded in the L category of band 3 appears; the number of times each symbol to be coded in the H category of band 3 appears together with the number of times the same symbol to be coded in the L category of band 4 appears; and for the H category of band 4, the number of times only that category appears. Based on this codebook 151 (especially the number of occurrences), the codeword to encode each symbol to be coded can be determined.

In the case of this codebook 151, the encoding control unit 15 assigns four encoders to the encoding unit 16 so as to create a configuration (second configuration) in which one encoder (common encoder) is assigned to the L categories of bands 1 and 2, one encoder is assigned to the H category of band 2 and the L category of band 3, one encoder is assigned to the H category of band 3 and the L category of band 4, and one encoder is assigned to the H category of band 4. The encoding control unit 15 then causes the encoding unit of the second configuration to encode the target symbol sequence based on the codeword. The encoding control unit 15 also determines the size of the compressed data (second compressed data) containing the encoded data obtained when encoding is performed in the second configuration and additional information for identifying the codeword. In the second configuration, the number of encoders can be reduced compared to the first configuration, thereby reducing the amount of additional information data.

The encoding control unit 15 determines which of the first compressed data in the first configuration and the second compressed data in the second configuration has a higher compression rate (e.g., which has a smaller data volume), and controls the compressed data output unit 18 to send the compressed data in the configuration with the higher compression rate to the decoder 30.

The encoding unit 16 performs a process of encoding the prediction difference under the control of the encoding control unit 15. The encoding unit 16 encodes δ ^· ₄ using a known technique, for example, first-order Markov compression as described in Non-Patent Document 1. In this embodiment, the encoding unit 16 also encodes information indicating the positive or negative sign of each symbol to be encoded using a known technique. The encoding unit 16 may also encode additional information using a known technique.

In addition, for {α ^· ₁ , β ^· ₁ , γ ^· ₁ }, {α ^· ₂ , β ^· ₂ , γ ^· ₂ }, {α ^· ₃ , β ^· ₃ , γ ^· ₃ }, and {α ^· ₄ , β ^· ₄ , γ ^· ₄ }, the encoding unit 16 performs a process of encoding the prediction difference using the first configuration and/or the second configuration in accordance with control by the encoding control unit 15.

Figure 12 is a diagram illustrating a first example configuration of an encoding unit according to one embodiment.

The encoding unit 16 includes an encoding unit 16-1-α that encodes {α ^· ₁ , α ^· ₂ , α ^· ₃ , α ^· ₄ }, an encoding unit 16-1-β that encodes {β ^· ₁ , β ^· ₂ , β ^· ₃ , β ^· ₄ }, and an encoding unit 16-1-γ that encodes {γ ^· ₁ , γ ^· ₂ , γ ^· ₃ , γ ^· ₄ }.

The encoding unit 16-1-α includes branching units 160a to 160c and encoders 161a to 161g. The encoding units 16-1-β and 16-1-γ have the same configuration as the encoding unit 16-1-α, and are equivalent to the encoding unit 16-1-α in FIG. 12, with α replaced by β and γ.

The encoder 161a (encoder E11) receives the prediction residual α ^· ₁ of band 1, performs encoding processing, and generates and outputs encoded data E11(α ^· ₁ ). Note that the codewords used in encoding by the encoders 161a to 161g are determined based on the number of occurrences in the codebook 150.

The branching unit 160a inputs α ^· ₂ , and inputs the encoding target symbol α ^· _{2 L} corresponding to the L category to the encoder 161b, and inputs the encoding target symbol α ^· _{2 H} corresponding to the H category to the encoder 161c.

The encoder 161b (encoder E12) receives as input the encoding target symbol α ^· _2L of the L category in band 2, encodes it, and generates and outputs encoded data E12 (α ^· _2L ).

The encoder 161c (encoder E13) receives the encoding target symbol α ^· _2H of the H category in band 2 and encodes it to generate and output encoded data E13 (α ^· _2H ).

The branching unit 160b inputs α ^· ₃ , and inputs the encoding target symbol α ^· _{3 L} corresponding to the L category to the encoder 161d, and inputs the encoding target symbol α ^· _{3 H} corresponding to the H category to the encoder 161e.

The encoder 161d (encoder E14) receives as input the encoding target symbol α ^· _3L of the L category in band 3, encodes it, and generates and outputs encoded data E14 (α ^· _3L ).

The encoder 161e (encoder E15) receives the encoding target symbol α ^· _3H of the H category in band 3 and encodes it to generate and output encoded data E15 (α ^· _3H ).

The branching unit 160c inputs α ^· ₄ , and inputs the encoding target symbol α ^· _{4 L} corresponding to the L category to the encoder 161f, and inputs the encoding target symbol α ^· _{4 H} corresponding to the H category to the encoder 161g.

The encoder 161f (encoder E16) receives as input the encoding target symbol α ^· _4L of the L category in band 4, encodes it, and generates and outputs encoded data E16 (α ^· _4L ).

The encoder 161g (encoder E17) receives the encoding target symbol α ^· _4H of the H category in band 4 and encodes it to generate and output encoded data E17 (α ^· _4H ).

Figure 13 is a diagram illustrating a second example configuration of an encoding unit according to one embodiment.

The encoding unit 16 includes an encoding unit 16-2-α that encodes {α ^· ₁ , α ^· ₂ , α ^· ₃ , α ^· ₄ }, an encoding unit 16-2-β that encodes {β ^· ₁ , β ^· ₂ , β ^· ₃ , β ^· ₄ }, and an encoding unit 16-2-γ that encodes {γ ^· ₁ , γ ^· ₂ , γ ^· ₃ , γ ^· ₄ }.

The encoding unit 16-2-α includes branching units 162a to 162c and encoders 163a to 163d. The encoding units 16-2-β and 16-2-γ have the same configuration as the encoding unit 16-2-α, and are equivalent to the encoding unit 16-2-α in FIG. 13, with α replaced by β and γ.

The branching unit 162a receives α ^· ₂ as input, and inputs the encoding target symbol α ^· _{2 L} corresponding to the L category to the encoder 163a, and inputs the encoding target symbol α ^· _{2 H} corresponding to the H category to the encoder 163b.

The encoder 163a (encoder E21) is an example of a common encoder, which inputs a prediction residual α ^· ₁ of band 1, performs encoding processing to generate and output encoded data E21(α ^· ₁ ), and inputs a symbol to be encoded α ^· _2L of L category of band 2, performs encoding to generate and output encoded data E21(α ^· _2L ). Note that the codewords used in encoding in the encoders 163a to 163d are determined based on the number of occurrences in the codebook 151.

The branching unit 162b inputs α ^· ₃ , and inputs the encoding target symbol α ^· _{3 L} corresponding to the L category to the encoder 163b, and inputs the encoding target symbol α ^· _{3 H} corresponding to the H category to the encoder 163c.

Encoder 163b (encoder E22) is an example of a common encoder, which inputs a symbol to be coded α ^· _2H corresponding to the H category of band 2, performs coding processing to generate and output coded data E22(α ^· _2H ), and inputs a symbol to be coded α ^· _3L of the L category of band 3, performs coding to generate and output coded data E22(α ^· _3L ). Note that the coded data E21(α ^· _2L ) output from encoder 163a and the coded data E22(α ^· _2H ) output from encoder 163b are concatenated in the order in which they were generated.

The branching unit 162c receives α ^· ₄ as input, and inputs the encoding target symbol α ^· _{4 L} corresponding to the L category to the encoder 163c, and inputs the encoding target symbol α ^· _{4 H} corresponding to the H category to the encoder 163d.

Encoder 163c (encoder E23) is an example of a common encoder, which inputs a symbol to be coded α ^· _3H corresponding to the H category of band 3, performs coding processing to generate and output coded data E23(α ^· _3H ), and inputs a symbol to be coded α ^· _4L of the L category of band 4, performs coding to generate and output coded data E23(α ^· _4L ). Note that coded data E22(α ^{·3L) output from encoder 163b and coded data E23(α·} _3H ⁾ _output from encoder 163c are concatenated in the order in which they were generated.

The encoder 163d (encoder E24) is an example of a category encoder, which inputs the encoding target symbol α ^· _4H corresponding to the H category of band 4, performs encoding processing, and generates and outputs encoded data E24(α ^· _4H ). Note that the encoded data E23(α ^· _4L ) output from the encoder 163c and the encoded data E24(α ^· _4H ) output from the encoder 163d are concatenated in the order in which they were generated.

Returning to the explanation of Fig. 4, the compressed data output unit 18, under the control of the encoding control unit 15, transmits the compressed data having a higher compression ratio out of the first compressed data in the first configuration and the second compressed data in the second configuration to the decoder 30. Here, in this embodiment, the compressed data includes encoded data obtained by encoding {α ^· ₁ , β ^· ₁ , γ ^· ₁ }, {α ^· ₂ , β ^· ₂ , γ ^· ₂ }, {α ^· ₃ , β ^· ₃ , γ ^· ₃ }, and {α ^· ₄ , β ^· ₄ , γ ^· ₄ }, encoded data obtained by encoding δ ^· ₄ , positive/negative sign data indicating the positive/negative sign of the symbol to be encoded, additional information for creating a codebook corresponding to the configuration used for encoding or data obtained by compressing that additional information, and information specifying the configuration that created the compressed data with the higher compression ratio (configuration specifying information).

The decoder 30 includes a compressed data input unit 31 (an example of a data receiving unit), a decoding control unit 32, a decoding unit 33, a reproduction unit 34, and an image display unit 35.

The compressed data input unit 31 inputs compressed data transmitted from the encoder 10 via the network 50.

The decoding control unit 32 controls the processing in the decoding unit 33 to perform decoding processing of the coded data coded by the encoder 10 and generate quantized prediction residuals {α ^· ₁ , β ^· ₁ , γ ^· ₁ }, {α ^· ₂ , β ^· ₂ , γ ^· ₂ }, {α ^· ₃ , β ^· ₃ , γ ^· ₃ }, {α ^· ₄ , β ^· ₄ , γ ^· ₄ }, δ ^· ₄ . The decoding control unit 32 specifies the configuration of the coding unit specified by configuration specifying information included in the compressed data, and assigns a decoder corresponding to the configuration to the decoding unit 33 to execute the decoding processing.

The decoding control unit 32 generates a codebook 150 or 151 for use in decoding {α ^· ₁ , β ^· ₁ , γ ^· ₁ }, {α ^· ₂ , β ^· ₂ , γ ^· ₂ }, {α ^· ₃ , β ^· ₃ , γ ^· ₃ }, {α ^· ₄ , β ^· ₄ , γ ^· ₄ }, based on the additional information. This allows the decoding control unit 32 to identify the correspondence between the codewords and symbols of the encoded data contained in the compressed data, and controls the decoder of the decoding unit 33 to operate in accordance with this correspondence.

The decoding control unit 32 controls the process of decoding δ ^· ₄ by a decoding process corresponding to the encoding of δ ^· ₄ in the encoder 10. The decoding control unit 32 also controls the process of decoding the positive and negative signs by a decoding process corresponding to the encoding of the positive and negative signs in the encoder 10.

The decoding unit 33 performs processing to decode the encoded data according to the control of the decoding control unit 32.

Figure 14 is a diagram illustrating a first example configuration of a decoding unit according to one embodiment. This example configuration is a configuration for decoding coded data encoded when the encoding unit 16 is in the first configuration.

The decoding unit 33 includes a decoding unit 33-1-α that decodes the coded data into {α ^· ₁ , α ^· ₂ , α ^· ₃ , α ^· ₄ }, a decoding unit 33-1-β that decodes the coded data into { ^{β ·} ₁ , β ^· ₂ , β· ₃ , β ^· ₄ }, and a decoding unit 33-1-γ that decodes the coded data into {γ ^· ₁ , γ ^· ₂ , γ ^· ₃ , γ ^· ₄ }.

Decoding unit 33-1-α includes branching units 330a to 330c and decoders 331a to 331g. Note that decoding unit 33-1-β and decoding unit 33-1-γ have the same configuration as decoding unit 33-1-α, and are equivalent to replacing α in FIG. 14 with β and γ.

The decoder 331a (decoder D11) receives the encoded data E11 (α ^· ₁ ), performs a decoding process, and generates and outputs α ^· ₁ . The code words in the encoded data in the decoders 331a to 331g can be determined based on the number of occurrences in the codebook 150.

The branching unit 330a receives E12(α ^· _2L ) and E13(α ^· _2H ), and inputs E12(α ^· _2L ) to a decoder 331b and E13(α ^· _2H ) to a decoder 331c. Whether it is E12(α ^· _2L ) or E13(α ^· _2H ) can be determined by the symbol decoded immediately before in the decoder 331b or 331c.

The decoder 331b (decoder D12) receives E12 (α ^· _2L ) and decodes it to generate and output α ^· _2L .

The decoder 331c (decoder D13) receives E13 (α ^· _2H ) and decodes it to generate and output α ^· _2H . Note that α ^· _2L output from the decoder 331b and α ^· _2H output from the decoder 331c are concatenated in the order in which they were generated to become α ^· ₂ .

The branching unit 330b receives E14(α ^· _3L ) and E15(α ^· _3H ), inputs E14(α ^· _3L ) to a decoder 331d, and inputs E15(α ^· _3H ) to a decoder 331e.

The decoder 331d (decoder D14) receives E14 (α ^· _3L ) and decodes it to generate and output α ^· _3L .

The decoder 331e (decoder D15) receives E15 (α ^· _3H ) and decodes it to generate and output α ^· _3H . Note that α ^· _3L output from the decoder 331d and α ^· _3H output from the decoder 331e are concatenated in the order in which they were generated to generate α ^· ₃ .

The branching unit 330c receives E16(α ^· _4L ) and E17(α ^· _4H ), and inputs E16(α ^· _4L ) to a decoder 331f and E17(α ^· _4H ) to a decoder 331g.

The decoder 331f (decoder D16) receives E16 (α ^· _4L ) and decodes it to generate and output α ^· _4L .

The decoder 331g (decoder D17) receives E17 (α ^· _4H ) and decodes it to generate and output α ^· _4H . Note that α ^· _4L output from the decoder 331f and α ^· _4H output from the decoder 331g are concatenated in the order in which they were generated to generate α ^· ₄ .

Figure 15 is a diagram illustrating a second configuration example of a decoding unit according to one embodiment. This configuration example is a configuration for decoding coded data when the encoding unit 16 is in the second configuration.

The decoding unit 33 includes a decoding unit 33-2-α that decodes the coded data into {α ^· ₁ , α ^· ₂ , α ^· ₃ , α ^· ₄ }, a decoding unit 33-2-β that decodes the coded data into { ^{β ·} ₁ , β ^· ₂ , β· ₃ , β ^· ₄ }, and a decoding unit 33-2-γ that decodes the coded data into {γ ^· ₁ , γ ^· ₂ , γ ^· ₃ , γ ^· ₄ }.

Decoding unit 33-2-α includes branching units 332a to 332c and decoders 333a to 333d. Here, decoders 333a to 333c are common decoders. Note that decoding units 33-2-β and 33-2-γ have the same configuration as decoding unit 33-2-α, and correspond to the case where α in FIG. 15 is replaced with β and γ.

The branching unit 332a receives E21(α ^· _2L ) and E22(α ^· _2H ), and inputs E21(α ^· _2L ) to the decoder 333a and E22(α ^· _2H ) to the decoder 333b. Whether it is E21(α ^· _2L ) or E22(α ^· _2H ) can be determined by the symbol decoded immediately before in the decoder 333a or 333b.

The decoder 333a (decoder D21) receives E21(α ^· ₁ ) and performs a decoding process to generate and output α ^· ₁ , and receives E21(α ^· ₂ ) and performs a decoding process to generate and output α ^· ₂ . The code words in the encoded data in the decoders 333a to 333d can be determined based on the number of occurrences in the codebook 151.

The branching unit 332b receives E22(α ^· _3L ) and E23(α ^· _3H ), and inputs E22(α ^· _3L ) to the decoder 333b and E23(α ^· _3H ) to the decoder 333c.

The decoder 333b (decoder D22) receives E22(α ^· _2H ) and performs a decoding process to generate and output α ^· _2H , and receives E22(α ^· _3L ) and performs a decoding process to generate and output α ^· _3L . Note that α ^· _2L output from the decoder 333a and α ^· _2H output from the decoder 333b are concatenated in the order in which they were generated to become α ^· ₂ .

The branching unit 332c receives E23(α ^· _4L ) and E24(α ^· _4H ), and inputs E23(α ^· _4L ) to the decoder 333c and E24(α ^· _4H ) to the decoder 333d.

The decoder 333c (decoder D23) receives E23 (α ^· _3H ) and performs a decoding process to generate and output α ^· _3H , and receives E23 (α ^· _4L ) and performs a decoding process to generate and output α ^· _4L . Note that α ^· _3L output from the decoder 333b and α ^· _3H output from the decoder 333c are concatenated in the order in which they were generated to form α ^· ₃ .

The decoder 333d (decoder D24) receives E24 (α ^· _4H ) and performs a decoding process to generate and output α ^· _4H . Note that α ^· _4L output from the decoder 333c and α ^· _4H output from the decoder 333d are concatenated in the order in which they were generated to form α ^· ₄ .

The reproduction unit 34 performs processing to reproduce the processing target data δ' 0 using the quantized prediction residuals {α ^· ₁ , β ^· ₁ , γ ^· ₁ }, {α ^· ₂ , β ^· ₂ , γ ^· ₂ }, {α ^· ₃ , β ^· ₃ , γ ^· ₃ }, {α ^· ₄ , β ^· ₄ , γ ^· ₄ }, δ ^· ₄ decoded by the decoding unit ₃₃ .

Figure 16 is a diagram showing the configuration of the reproduction unit according to one embodiment.

The reproduction unit 34 includes a partial reproduction unit 34-4, a partial reproduction unit 34-3, a partial reproduction unit 34-2, and a partial reproduction unit 34-1.

The partial reproduction unit 34-4 receives input of {α ^· ₄ , β ^· ₄ , γ ^· ₄ , δ ^· ₄ } and outputs a reproduction value δ' _3. The partial reproduction unit 34-3 receives input of {α ^· ₃ , β ^· ₃ , γ ^· ₃ } and δ' ₃ and outputs a reproduction value δ' _2. The partial reproduction unit 34-2 receives input of {α ^· ₂ , β ^· ₂ , γ ^· ₂ } and δ' ₂ and outputs a reproduction value δ' _1. The partial reproduction unit 34-1 receives input of {α ^· ₁ , β ^· ₁ , γ ^· ₁ } and δ' ₁ and outputs δ' ₀ .

Figure 17 is a diagram illustrating the configuration of one partial reproduction unit according to one embodiment. Figure 17 is a diagram illustrating the configuration of the partial reproduction unit 34-4.

The partial reconstruction unit 34-4 includes a DPCM encoder 341, an ACP (alternating current component prediction) unit 342, an inverse quantization unit 343, a calculation unit 344, and a partial subband synthesis unit 345.

The DPCM encoder 341 is capable of performing differential pulse code modulation processing, inputting δ ^· ₄ and outputting a reproduced value δ′ ₄ .

The ACP unit 342 receives the reproduced value δ′ ₄ and calculates {α ^∼ ₄ , β ^∼ ₄ , γ ^∼ ₄ } by AC component prediction. {α ^∼ ₄ , β ^∼ ₄ , γ ^∼ ₄ } are predicted values for {α ₄ , β ₄ , γ ₄ } predicted by AC component prediction.

The inverse quantization unit 343 performs inverse quantization on {α ^· ₄ , β ^· ₄ , γ ^· ₄ } and outputs the result.

The calculation unit 344 adds the predicted values {α ^∼ ₄ , β ^∼ ₄ , γ ^∼ ₄ } to the result of the inverse quantization by the inverse quantization unit 343 to calculate the reproduced values {α′ ₄ , β′ ₄ , γ′ ₄ }.

The partial subband synthesis unit 345 receives the reproduction values {α' ₄ , β' ₄ , γ' ₄ } and the reproduction value δ' ₄ , performs the inverse transformation of the subband division process, and outputs the reproduction value δ' ₃ .

Figure 18 is a diagram showing the configuration of another partial reproduction unit according to one embodiment. Figure 18 shows the configuration of a partial reproduction unit 34-n (where n is 1 to 3).

The partial reproduction unit 34-n includes an ACP unit 342, an inverse quantization unit 343, an arithmetic unit 344, and a partial subband synthesis unit 345. Note that the partial reproduction unit 34-n does not include a DPCM encoder 341 because it can obtain the reproduction value δ' _n from the lower-level partial reproduction unit 34-n+1.

The ACP unit 342 receives the reproduced value δ′ _n and calculates {α ^∼ _n , β ^∼ _n , γ ^∼ _n } by AC component prediction. {α ^∼ _n , β ^∼ _n , γ ^∼ _n } are predicted values for {α _n , β _n , γ _n } predicted by AC component prediction.

The inverse quantization unit 343 performs inverse quantization on {α ^· _{n 1} , β ^· _{n 1} , γ ^· _{n 1} } and outputs the result.

The calculation unit 344 adds the predicted values {α ^∼ _n , β ^∼ _n , γ ^∼ _n } to the result of the inverse quantization by the inverse quantization unit 343 to calculate the reproduced values {α' _n , β' _n , γ _{' n} }.

The partial subband synthesis unit 345 receives the reproduction values {α' _n , β' _n , γ' _n } and the reproduction value δ' _n , performs the inverse transformation of the subband decomposition process, and outputs a reproduction value δ' _n-1 . For example, the partial subband synthesis unit 345 of the partial reproduction unit 34-1 outputs a reproduction value δ' ₀ for the data to be processed.

Returning to the explanation of FIG. 4, the image display unit 35 displays an image based on the processing object data δ′ ₀ reproduced by the reproduction unit 34 .

Next, we will explain the processing operations in image processing system 1.

First, we will explain the encoding process performed by the encoder 10.

Figure 19 is a flowchart of the encoding process according to one embodiment.

In the encoding process, first, the target data input unit 11 inputs the image to be encoded (target data), and the subband decomposition unit 12 performs hierarchical subband decomposition of the target data (step S11).

Next, the prediction residual processing unit 14 outputs prediction differences {α _·1 , β _·1 , γ _·1 }, {α _·2 , β _·2 , γ _·2 }, {α _·3 , β _·3 , γ· ₃ }, {α _·4 , β _·4 , γ· ₄ }, δ· ₄ for each data after subband division {α ^· ₁ , β ^· ₁ , γ ^· ₁ }, {α ^· ₂ , β ^· ₂ , γ ^· ₂ }, {α ^· ₃ , β ^· ₃ , γ ^· ₃ }, {α ^· ₄ , β ^· ₄ , γ ^· ₄ }, δ ^· ₄ (step S12).

The encoding control unit 15 then determines a sequence of symbols to be encoded that describes each piece of data, and divides at least one band into multiple categories based on the energy of the symbol immediately preceding the symbol to be encoded. In this embodiment, the categories are divided for bands 2 to 4. The encoding control unit 15 then counts the number of times a symbol belonging to each category appears immediately preceding the symbol to be encoded, and creates a codebook 150. Next, based on the codebook 150, the encoding control unit 15 assigns the seven encoders of the encoding unit 16 so that one encoder is assigned to band 1, which is not divided into categories, and one encoder is assigned to each of the categories in bands 2 to 4 (first configuration) (step S13).

The encoding control unit 15 then causes the encoding unit 16 of the first configuration to encode the encoding target symbol sequence based on the codeword, and determines the size of the compressed data (first compressed data) that includes the encoded data obtained when encoding in the first configuration is performed and additional information for identifying the codeword (step S14).

Next, the encoding control unit 15 creates a codebook 151 that assumes a case where a band or its category and a closely related band (e.g., an adjacent band, or a band immediately above or below in the hierarchy) or category of that band are combined into one and a common encoder. Next, based on the codebook 151, the encoding control unit 15 assigns the encoders of the encoding unit 16 to a configuration (second configuration) in which one encoder (common encoder) is assigned to the L category of bands 1 and 2, one encoder is assigned to the H category of band 2 and the L category of band 3, one encoder is assigned to the H category of band 3 and the L category of band 4, and one encoder is assigned to the H category of band 4 (step S15).

The encoding control unit 15 then causes the second encoding unit 16 to encode the encoding target symbol sequence based on the codeword, and determines the size of the compressed data (second compressed data) that includes the encoded data obtained when encoding in the second configuration is performed and additional information for identifying the codeword (step S16).

Next, the encoding control unit 15 determines which of the first compressed data in the first configuration and the second compressed data in the second configuration has a higher compression rate (step S17).

As a result, if the compression ratio of the first compressed data is higher (step S17: first configuration), the encoding control unit 15 causes the compressed data output unit 18 to transmit the first compressed data to the decoder 30 (step S18). Note that the first compressed data includes other information necessary for decoding the data to be processed (e.g., encoded data obtained by encoding δ ^· ₄ , encoded data of the positive/negative sign of the symbol to be encoded, information capable of identifying the first configuration that performed the encoding (configuration identification information), etc.).

On the other hand, if the compression ratio of the second compressed data is higher (step S17: second configuration), the encoding control unit 15 causes the compressed data output unit 18 to transmit the second compressed data to the decoder 30 (step S19). Note that the second compressed data includes other information necessary for decoding the data to be processed (e.g., encoded data obtained by encoding δ ^· ₄ , encoded data of the symbol to be encoded, information capable of identifying the second configuration used for encoding (configuration identification information), etc.).

This encoding process transmits compressed data with a high compression rate, improving the compression efficiency of the compressed data sent to the decoder 30.

Next, we will explain the decoding process performed by the decoder 30.

Figure 20 is a flowchart of the decoding process according to one embodiment.

The compressed data input unit 31 inputs compressed data transmitted from the encoder 10 via the network 50 (step S31).

Next, the decoding control unit 32 determines whether the compressed data input by the compressed data input unit 31 is compressed data in the first configuration or the second configuration (step S32).

As a result, if the compressed data is first compressed data in the first configuration (step S32: first configuration), the decoding control unit 32 generates a codebook 150 based on the additional information and configures the decoding unit 33 to have a decoder configuration (first configuration) that can perform decoding corresponding to the encoding by the encoding unit 16 in the first configuration (step S33).

On the other hand, if the compressed data is second compressed data in the second configuration (step S32: second configuration), the decoding control unit 32 generates a codebook 151 based on the additional information and configures the decoding unit 33 to have a decoder configuration (second configuration) that can perform decoding corresponding to the encoding by the encoding unit 16 in the second configuration (step S34).

Next, the decoding control unit 32 decodes the coded data included in the compressed data using the configured decoding unit 33 (step S35), which outputs quantized prediction residuals {α ^· ₁ , β ^· ₁ , γ ^· ₁ }, {α ^· ₂ , β ^· ₂ , γ ^· ₂ }, {α ^· ₃ , β ^· ₃ , γ ^· ₃ }, {α ^· ₄ , β ^· ₄ , γ ^· ₄ }, δ ^· ₄ .

Next, the reproduction unit 34 performs a process of reproducing the processing target data _{δ' 0 using the quantized prediction residuals {α·1, β·1, γ·1}, {α·2, β·2 ,} ^γ _· ² _} ^, _{ ^{α ·} ₃ ^{, β} _· ^{3 ,} _γ ^· ₃ ^} _, ^{ _{α ·} 4, β·4, γ ^· ₄ }, δ· ₄ (step S36).

Next, the image display unit 35 displays an image based on the processing object data δ' ₀ reproduced by the reproduction unit 34 (step S37).

The above-described decoding process allows images to be displayed appropriately based on the compressed data created by the encoder 10.

The above-mentioned encoder 10 and decoder 30 can each be configured as a computer device.

Figure 21 is a diagram illustrating the configuration of a computer device according to one embodiment. Note that in this embodiment, the encoder 10 and decoder 30 are configured as separate computer devices, but these computer devices can have similar configurations. Therefore, for convenience, the following explanation will use the computer device shown in Figure 21 to describe the computer devices that make up the encoder 10 and decoder 30.

The computer device 100 includes, for example, a central processing unit (CPU) 101, main memory 102, a graphics processing unit (GPU) 103, a reader/writer 104, a communication interface (communication I/F) 105, an auxiliary storage device 106, an input/output interface (input/output I/F) 107, a display device 108, and an input device 109. The CPU 101, main memory 102, GPU 103, reader/writer 104, communication I/F 105, auxiliary storage device 106, input/output I/F 107, and display device 108 are connected via a bus 110. The encoder 10 and decoder 30 are each configured by appropriately selecting some or all of the components described in the computer device 100.

In the computer device 100 that constitutes the encoder 10, the CPU 101 executes a data processing program stored in the auxiliary storage device 106 to constitute, for example, a target data input unit 11, a subband division unit 12, a prediction residual processing unit 14, an encoding control unit 15, an encoding unit 16, and a compressed data output unit 18.

In the computer device 100 that constitutes the decoder 30, the CPU 101 executes a data processing program stored in the auxiliary storage device 106 to constitute, for example, a compressed data input unit 31, a decoding control unit 32, a decoding unit 33, a reproduction unit 34, and an image display unit 35.

The main memory 102 is, for example, RAM, ROM, etc., and stores programs (data processing programs, etc.) executed by the CPU 101 and various information. The auxiliary storage device 106 is, for example, a non-transitory storage device (non-volatile storage device) such as a hard disk drive (HDD) or solid state drive (SSD), and stores programs executed by the CPU 101 and various information.

The GPU 103 is a processor that is suitable for executing specific processes, such as image processing, and is suitable for executing processes that are performed in parallel, for example. In this embodiment, the GPU 103 executes predetermined processes in accordance with instructions from the CPU 101.

The reader/writer 104 is detachable from the recording medium 111 and reads and writes data to the recording medium 111. Examples of the recording medium 111 include non-transitory recording media (non-volatile recording media) such as SD memory cards, floppy disks (FDs, registered trademark), CDs, DVDs, BDs (registered trademarks), and flash memory. In this embodiment, a data processing program may be stored on the recording medium 111 and read and used by the reader/writer 104. Furthermore, in the computer device 100 constituting the encoder 10, image data to be processed may be stored on the recording medium 111 and read and used by the reader/writer 104. Furthermore, in the computer device 100 constituting the encoder 10, the reader/writer 104 may store compressed data on the recording medium 111. Furthermore, in the computer device 100 constituting the decoder 30, the reader/writer 104 may read and use compressed data from the recording medium 111.

The communication I/F 105 is connected to the network 50 and transmits and receives data to and from other devices connected to the network 50.

The input/output I/F 107 is connected to an input device 109, such as a mouse or keyboard. In the computer device 100 that constitutes the encoder 10, the input/output I/F 107 accepts operational input from a user of the encoder 10 using the input device 109. In addition, in the computer device 100 that constitutes the decoder 30, the input/output I/F 107 accepts operational input from a user of the decoder 30 using the input device 109.

The display device 108 is, for example, a display device such as an LCD display, and displays and outputs various information. The display device 108 is used, for example, in the decoder 30, to display images using the image display unit 35.

The present invention is not limited to the above-described embodiments and modifications, and can be implemented with appropriate modifications within the scope of the spirit of the present invention.

For example, in the above embodiment, the encoding unit 16 is configured as a first configuration and a second configuration, and a determination is made as to which configuration has a higher compression rate based on the amount of compressed data when each configuration is used, and data encoded using the determined configuration is then transmitted to the decoder 30. However, the present invention is not limited to this. For example, when processing data that is expected to have a higher compression rate when encoded using the second configuration, encoding may always be performed using the second configuration.

In addition, in the above embodiment, the compression rate of the processing target data to be actually transmitted to the decoder 30 was determined based on the amount of compressed data when the encoding unit 16 was configured in either the first or second configuration. However, the present invention is not limited to this. For example, it is possible to prepare determination data similar to the processing target data in advance, and use this determination data to determine which configuration has a higher compression rate based on the amount of compressed data when the encoding unit 16 is configured in either the first or second configuration, and then encode the processing target data to be transmitted using the resulting determined configuration.

Furthermore, in the above embodiment, an example was shown in which the computer device 100 was equipped with either the encoder 10 or the decoder 30. The present invention is not limited to this, and the computer device 100 may be equipped with both the encoder 10 and the decoder 30.

1...Data processing system, 10...Encoder, 11...Target data input unit, 12...Subband decomposition unit, 14...Prediction residual processing unit, 15...Encoding control unit, 16...Encoding unit, 18...Compressed data output unit, 30...Decoder, 31...Compressed data input unit, 32...Decoding control unit, 33...Decoding unit, 34...Reproduction unit, 35...Image display unit, 50...Network, 100...Computer device, 101...CPU, 102...Main memory, 108...Display device, 111...Recording medium, 161a-161g, 163a-163d...Encoders, 331a-331g, 333a-333d...Decoders

Claims (8)

An encoder that compresses predetermined processing target data,
a frequency band dividing unit that divides data to be processed into a plurality of partial frequency bands;
a category dividing unit that divides, for at least some partial frequency bands, a plurality of encoding target symbols included in a compression target element based on elements of the partial frequency band into a plurality of categories based on the energy of a symbol immediately preceding the encoding target symbol in a sequence of the encoding target symbols;
an encoding unit having a plurality of encoders that encode encoding target symbols included in compression target elements based on elements of the plurality of partial frequency bands;
At least one of the encoders is a common encoder that encodes, using a common code, the encoding target symbols of the elements to be compressed included in one of the categories and the encoding target symbols of the elements to be compressed in a partial frequency band adjacent to the partial frequency band to which the one of the categories belongs . At least one of the common encoders comprises:
2. The encoder according to claim 1, further comprising a common encoder that encodes, using a common code, symbols to be encoded of elements to be compressed included in one of the categories and symbols to be encoded of elements to be compressed in categories that belong to a partial frequency band adjacent to the partial frequency band to which the one of the categories belongs . The plurality of encoders
2. The encoder according to claim 1, further comprising: a category encoder for encoding a symbol to be encoded of each of the elements to be compressed in the category that is not encoded by the common encoder. The plurality of encoders
a first configuration not including the common coder and a second configuration including at least one common coder;
a determination unit that determines whether the first configuration or the second configuration has a higher compression ratio based on a data amount of first compressed data, the first compressed data including data obtained by encoding predetermined determination data using the first configuration and additional information that allows codebooks for the plurality of encoders of the first configuration to be generated, and a data amount of second compressed data, the second compressed data including data obtained by encoding predetermined determination data using the second configuration and additional information that allows codebooks for the plurality of encoders of the second configuration to be generated,
The encoder according to claim 3 , wherein the encoding unit encodes the data to be processed using a configuration determined by the determination unit to have a high compression ratio. A data processing method using an encoder that compresses predetermined processing target data, comprising:
Dividing the data to be processed into a plurality of partial frequency bands;
for at least a portion of partial frequency bands, dividing a plurality of encoding target symbols included in a compression target element based on elements of the partial frequency band into a plurality of categories based on the energy of a symbol immediately preceding the encoding target symbol in a sequence of the encoding target symbols;
Encoding encoding target symbols of compression target elements based on elements of the plurality of partial frequency bands by a plurality of encoders;
A data processing method in which at least one of the encoders is a common encoder that encodes, using a common code, symbols to be encoded of elements to be compressed included in one of the categories and symbols to be encoded of elements to be compressed in a partial frequency band adjacent to the partial frequency band to which the one of the categories belongs . A data processing program executed by a computer constituting an encoder that compresses predetermined processing target data,
The computer
a frequency band dividing unit that divides data to be processed into a plurality of partial frequency bands;
a category dividing unit that divides, for at least some partial frequency bands, a plurality of encoding target symbols included in a compression target element based on elements of the partial frequency band into a plurality of categories based on the energy of a symbol immediately preceding the encoding target symbol in a sequence of the encoding target symbols;
an encoding unit including a plurality of encoders for encoding encoding target symbols of compression target elements based on at least some elements of the plurality of partial frequency bands;
A data processing program in which at least one of the encoders is a common encoder that encodes, using a common code, symbols to be encoded of elements to be compressed included in one of the categories and symbols to be encoded of elements to be compressed in a partial frequency band adjacent to the partial frequency band to which the one of the categories belongs . 1. A data processing system including an encoder that compresses predetermined processing target data, and a decoder that decodes the data compressed by the encoder,
The encoder comprises:
a frequency band dividing unit that divides data to be processed into a plurality of partial frequency bands;
a category dividing unit that divides, for at least some partial frequency bands, a plurality of encoding target symbols included in a compression target element based on elements of the partial frequency band into a plurality of categories based on the energy of a symbol immediately preceding the encoding target symbol in a sequence of the encoding target symbols;
an encoding unit having a plurality of encoders that encode encoding target symbols of compression target elements based on at least some elements of the plurality of partial frequency bands;
a data transmitting unit configured to transmit, to the decoder, encoded data including the encoded encoding target symbol and additional information that enables the plurality of encoders to generate a codebook that can identify a code obtained by encoding;
and
at least one of the encoders is a common encoder that encodes, using a common code, encoding symbols to be encoded of elements to be compressed included in one of the categories and symbols to be encoded of elements to be compressed in a partial frequency band adjacent to a partial frequency band to which the one of the categories belongs;
The decoder
a data receiving unit that receives the compressed data;
a decoding unit having a plurality of decoders that decode the encoded encoding target symbols based on the additional information included in the compressed data;
a reconstructing unit that reconstructs the processing target data based on the encoding target symbols of the partial frequency band decoded by the decoding unit;
and
A data processing system, wherein at least one of the decoders is a common decoder that decodes symbols to be encoded that have been encoded by the common encoder. A decoder for decoding compressed data, comprising:
a data receiving unit that receives compressed data including encoded data of the data to be processed and additional information that can generate a codebook that can identify a code obtained by encoding;
a decoding unit having a plurality of decoders that decode the encoded encoding target symbols based on the additional information;
a reproducing unit that reproduces the data to be processed based on the encoding target symbols of the compression target elements of the partial frequency band decoded by the decoding unit;
and
At least one of the decoders is a common decoder that decodes, for at least a portion of the partial frequency bands of the data to be processed, coded data in which a symbol to be coded included in one of a plurality of divided categories and a symbol to be coded of an element to be compressed in a partial frequency band adjacent to the partial frequency band to which the one category belongs are coded using a common code, based on the energy of the immediately preceding symbol in the sequence of the symbol to be coded.