JP4438655B2 - Encoding device, decoding device, encoding method, and decoding method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To enable efficient vector quantization without using a previously prepared vector quantization table. <P>SOLUTION: In a coder 100, a framing section 1 divides an input signal into frames. A frequency conversion section 2 applies frequency conversion to the input signal per unit (block) constituting one frame. A coding section 3 applies coding to a frequency conversion coefficient for one unit in the one frame by using a predetermined coding system. A vector quantizing section 4 generates a vector quantization table on the basis of the coded frequency conversion coefficient for one unit, and applies vector quantization to the other frequency conversion coefficient in the same frame by using the generated vector quantization table. <P>COPYRIGHT: (C)2006,JPO&amp;NCIPI

Description

  The present invention relates to an encoding device that encodes a signal, a decoding device that decodes an encoded signal, an encoding method, and a decoding method.

In vector quantization used as a compression method for audio signals, image signals, etc., generally, a representative vector that is most similar to an input signal is selected from a vector quantization table in which a plurality of representative vectors are stored in advance. The index indicated by the selected representative vector is output as a code (see, for example, Patent Document 1).
Japanese Patent No. 2683734

  In conventional vector quantization, in order to perform efficient encoding, it is necessary to collect a large amount of sample data (representative vectors) expected to be input and generate a vector quantization table in advance. Therefore, the vector quantization table cannot be easily generated.

  An object of the present invention is to enable efficient vector quantization without using a vector quantization table prepared in advance.

In order to solve the above-described problem, the encoding device according to claim 1 performs framing means for dividing an input signal into frames, and performs frequency conversion on the input signal for each block unit constituting one frame. Based on the frequency conversion means , the encoding means for encoding the frequency conversion coefficient for a predetermined block in one frame, and the frequency conversion coefficient for one block that is the object of encoding by the encoding means Using a generation unit that generates a vector quantization table and a vector quantization table generated by the generation unit, a frequency conversion coefficient of another block in the same frame that is not encoded by the encoding unit And vector quantization means for performing vector quantization.

According to a second aspect of the present invention, in the encoding device according to the first aspect, a calculation means for calculating a frequency distribution of frequency transform coefficient values for each block unit constituting one frame, and a calculation by the calculation means Selecting means for selecting a frequency transform coefficient for one block to be used for generating a vector quantization table based on the frequency distribution, and the encoding means has a frequency transform coefficient for the predetermined one block. as is characterized by using frequency transform coefficients for one block selected by said selection means.

According to a third aspect of the present invention, in the encoding device according to the first aspect, the encoding means uses a frequency conversion coefficient of the first block of one frame as the frequency conversion coefficient for the predetermined one block. It is characterized by that.

The decoding device according to claim 4 is a decoding device that decodes an input signal encoded by the encoding device according to claim 1, and is encoded by the encoding means of the encoding device. Decoding means for decoding frequency transform coefficients for a predetermined block in a frame ; generation means for generating a vector quantization table based on the frequency transform coefficients for one block decoded by the decoding means; and the encoding A signal obtained by performing vector quantization on the frequency transform coefficients of other blocks in the same frame by the vector quantization means of the apparatus is subjected to inverse vector quantization using the vector quantization table generated by the generation means. Inverse vector quantization means for performing frequency inverse transform means for performing frequency inverse transform on the frequency transform coefficient obtained by the inverse vector quantization It is characterized by a door.

The encoding method according to claim 5 includes a framing step of dividing an input signal into frames, a frequency conversion step of performing frequency conversion on the input signal for each block unit constituting one frame, and 1 frame. A vector quantization table is generated based on an encoding step for encoding a predetermined one block of frequency transform coefficients and a frequency transform coefficient for one block to be encoded in the encoding step. A vector that performs vector quantization on the frequency transform coefficients of other blocks in the same frame that are not encoded in the encoding step using the generation step and the vector quantization table generated in the generation step And a quantization step .

A decoding method according to claim 6 is a decoding method for decoding an input signal encoded by the encoding method according to claim 5, and is encoded by the encoding step of the encoding method. A decoding step of decoding frequency transform coefficients for a predetermined block in a frame; a generation step of generating a vector quantization table based on the frequency transform coefficients of one block decoded by the decoding step; and the encoding A signal obtained by performing vector quantization on frequency transform coefficients of other blocks in the same frame by the vector quantization step of the method is subjected to inverse vector quantization using the vector quantization table generated by the generation step. Performing inverse vector quantization step, and applying frequency inverse transform to the frequency transform coefficient obtained by the inverse vector quantization It is characterized and the wave number inverse conversion step, in that it consists of.

  According to the present invention, a vector quantization table is generated from one unit of frequency conversion coefficients in one frame, and a vector quantization table is generated for the frequency conversion coefficients to be processed using the generated vector quantization table. Thus, efficient vector quantization can be performed without preparing a vector quantization table that collects a large amount of sample data (representative vectors) as in the prior art.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
(Embodiment 1)

A first embodiment of the present invention will be described with reference to FIGS.
First, the configuration in the first embodiment will be described.

  FIG. 1A shows a configuration of an encoding apparatus 100 according to Embodiment 1 of the present invention. As illustrated in FIG. 1A, the encoding device 100 includes a framing unit 1, a frequency conversion unit 2, an encoding unit 3, and a vector quantization unit 4.

  The framing unit 1 divides the input signal into fixed-length frames that are processing units for encoding (compression), and outputs the frames to the frequency converting unit 2. One frame is composed of a plurality of blocks which are units for performing frequency conversion.

The frequency conversion unit 2 performs frequency conversion on the signal input from the framing unit 1 in units of blocks and outputs the result to the encoding unit 3. As the frequency conversion, for example, DFT (Discrete Fourier Transform) can be used. If the input signal is {x _n | n = 0,..., N−1}, the frequency conversion coefficients {X _k | k = 0,..., N−1} in the DFT are defined as in equation (1). The

  The encoding unit 3 encodes one block of frequency conversion coefficients in a predetermined encoding method among the frequency-converted blocks included in one frame. Here, various coding methods such as Huffman coding and Range Coder coding can be used as the coding method. The encoding target in the encoding unit 3 may be a preset block such as the first block on the time axis in one frame, for example.

  The vector quantization unit 4 generates a vector quantization table from the frequency transform coefficients for one block encoded by the encoding unit 3, and uses the generated vector quantization table to generate other blocks in the same frame. Vector quantization is performed on each frequency transform coefficient for each block, and the signal subjected to vector quantization and the signal for one block encoded by the encoding unit 3 are output as encoded signals.

FIG. 2 shows an example of a vector quantization table 31 having a vector length of 3. Assuming that the frequency transform coefficient for one block encoded by the encoding unit 3 is {X _k | k = 0,..., N−1}, the vector quantization table 31 has three consecutive values for one index. There are N-2 indexes corresponding to frequency conversion coefficients.

なお、式（２）に示す誤差ｅ_jが最小となるインデックスｊを符号化信号として出力してもよいし、これをハフマン符号化等により更に符号化した後で出力するようにしてもよい。 Hereinafter, vector quantization using a vector quantization table having a vector length of 3 will be described as an example of vector quantization. Let {s _i | i = 0,1,2} be the frequency transform coefficient to be vector-quantized, and {V _j | j = 0,..., N−3} be the vector quantization table. Here, V _j = {v _ji | i = 0,1,2}. In this case, a vector of the vector quantization target (frequency conversion coefficient), a code for outputting the index j such error e _j is minimized for each element in the j-th vector of the vector quantization table. Error e _j is calculated, for example, as in Equation (2).

Incidentally, it may be the indices j that error e _j shown in equation (2) is minimized as a coded signal, which may be outputted after further encoded by Huffman coding, or the like.

  FIG. 1B shows a configuration of a decoding device 200 according to the embodiment of the present invention. The decoding device 200 is a device that decodes the signal encoded by the encoding device 100. As shown in FIG. 1B, the decoding device 5, the inverse vector quantization 6, the frequency inverse transformation unit 7, the frame synthesis. It is comprised by the part 8.

  The decoding unit 5 decodes the encoded signal for one block (encoded by the encoding unit 3) that is first encoded on the encoding device 100 side, and outputs the decoding result to the inverse vector quantization unit 6 To do. In addition, in the encoding apparatus 100, when encoding such as Huffman encoding is further performed after vector quantization of other blocks in the same frame, the decoding unit 5 also outputs encoded signals of the other blocks. Decrypt.

  The inverse vector quantization unit 6 generates a vector quantization table from the frequency transform coefficients for one block decoded by the decoding unit 5, and uses the generated vector quantization table to determine other blocks in the same frame. The signal is subjected to inverse vector quantization for each block, and the signal after inverse vector quantization is output to the frequency inverse transform unit 7. Here, the generation method of the vector quantization table is the same as that of the encoding apparatus 100. In this inverse vector quantization, a vector corresponding to the input index is extracted from the vector quantization table.

  The frequency inverse transform unit 7 performs frequency inverse transform on the signal input from the inverse vector quantization unit 6 and outputs the result to the frame synthesis unit 8.

  The frame synthesizing unit 8 synthesizes frames that are processing units of encoding and decoding, and outputs the synthesized signal as a reproduction signal.

Next, the operation in the first embodiment will be described.
First, with reference to the flowchart of FIG. 3, the encoding process performed in the encoding apparatus 100 of Embodiment 1 is demonstrated.

  First, the input signal is divided into frames (step S1), and the input signal is subjected to frequency conversion in units of blocks constituting one frame (step S2). Next, the frequency transform coefficient for one block in one frame (for example, the first block on the time axis) is encoded by a predetermined encoding method (for example, Huffman encoding) (step S3), and the encoding is performed. A vector quantization table (see FIG. 2) is generated from the frequency transform coefficients of the blocks (step S4).

  Next, it is determined whether or not encoding has been performed for all the blocks in one frame (step S5). When it is determined in step S5 that encoding of all blocks in one frame has not been completed (step S5; NO), the frequency conversion coefficient for one block to be encoded is generated in step S4. Vector quantization is performed using the vector quantization table, and further, encoding is performed using a predetermined encoding method (for example, Huffman encoding) (step S6).

  When the vector quantization and encoding for one block in step S6 are completed, the process returns to step S5, and the process of step S6 is repeated for blocks that have not been encoded yet. If it is determined in step S5 that encoding of all blocks in one frame has been completed (step S5; YES), the encoding process ends. Note that the processing in steps S2 to S6 is performed for each frame.

  Next, a decoding process executed in the decoding device 200 according to the first embodiment will be described with reference to the flowchart in FIG.

  First, the encoded signal for one block first encoded on the encoding apparatus 100 side is decoded (step T1), and a vector quantization table is generated from the frequency transform coefficients for one block obtained by the decoding. (Step T2).

  Next, it is determined whether or not the encoded signals of all the blocks in one frame have been decoded (step T3). If it is determined in step T3 that decoding of all blocks in one frame has not been completed (step T3; NO), an encoded signal of one block that has not yet been decoded in the frame is decoded. Inverse vector quantization is performed using the vector quantization table generated in step T2 (step T4).

  When decoding and inverse vector quantization for one block in step T4 are completed, the process returns to step T3, and the process of step T4 is repeated for the blocks that have not been decoded yet. If it is determined in step T3 that decoding of all blocks in one frame has been completed (step T3; YES), frequency inverse transform is performed on the signals obtained in steps T1 and T4 (step T5). The decryption process ends. This decoding process is performed for each frame, and the decoded frames are combined and output as a reproduction signal.

  As described above, according to the first embodiment of the present invention, the vector quantization table is generated from the frequency conversion coefficient of a predetermined block (for example, the first block on the time axis) in one frame, and the generated vector quantization table is generated. A vector that collects a large amount of sample data (representative vectors) as before, by performing vector quantization on the frequency conversion coefficients of other blocks in the same frame using the vector quantization table. Efficient vector quantization is possible without preparing a quantization table in advance. In particular, in audio signals and image signals, temporally continuous signals are often similar, so that efficient vector quantization is possible with a small amount of indexes.

  In addition, although the case where DFT was used as frequency conversion was shown in Embodiment 1, the frequency conversion applied in the frequency conversion part 2 is not limited to this. For example, DCT (Discrete Cosine Transform), MDCT (Modified Discrete Cosine Transform), etc. are applicable. In general, DCT is often used for image signals, and MDCT is often used for audio signals.

When DCT is used as frequency conversion, if the input signal is {x ₁ | l = 0,..., N−1}, the frequency conversion coefficient {X _m | m = 0,. It is defined as equation (3).

ここで、ｈ_nは窓関数であり、式（５）のように定義される。

When MDCT is used for frequency conversion, if the length of the MDCT block (each block constituting one frame) is M and the input signal is {x _n | n = 0,. The frequency conversion coefficient {X _k | k = 0,..., M / 2-1} to be obtained is defined as shown in Equation (4).

Here, h _n is a window function and is defined as in equation (5).

In the first embodiment, the block used for generating the vector quantization table has been determined in advance. However, the block having the largest number of types of frequency transform coefficients included in one block is designated. You may do it. For this, a histogram (frequency distribution) of frequency transform coefficient values is calculated for each block, and a block having the largest frequency transform coefficient value that appears one or more times is selected as a block used for generating a vector quantization table. Thereby, a vector quantization table with more patterns can be obtained, and the encoding accuracy can be improved.
(Embodiment 2)

A second embodiment of the present invention will be described with reference to FIGS.
In the first embodiment, the case where vector quantization is performed on other blocks in the same frame using the vector quantization table generated from the blocks in the same frame has been described. A case where vector quantization using the vector quantization table generated in the frame is possible will be described.

First, the configuration in the second embodiment will be described.
The configurations of the encoding device and the decoding device according to the second embodiment are the same as those of the encoding device 100 and the decoding device 200 shown in FIG. Hereinafter, only differences from the encoding apparatus 100 and the decoding apparatus 200 according to the first embodiment will be described.

  Similarly to the first embodiment, the vector quantization unit 4 in the encoding device 100 according to the second embodiment generates a vector quantization table from the frequency transform coefficients for one block encoded by the encoding unit 3. Using the vector quantization table, vector quantization is performed for each block on the frequency transform coefficients of other blocks in the same frame.

Then, the vector quantization unit 4 calculates a vector quantization distortion based on the vector quantization result. This vector quantization distortion can be expressed by, for example, an average of error energies between a value obtained by vector quantization and an original value before vector quantization. A set of vectors of vector length k is {S _j | j = 0,..., N−1} and S _j = {s _ji | i = 0,..., K−1}, and these are obtained by vector quantization. Let {S ′ _j | j = 0,..., N−1} and S ′ _j = {s ′ _ji | i = 0 _,. The vector quantization distortion e at this time is expressed as shown in Equation (6).

  Next, the vector quantization unit 4 uses the vector quantization table generated in the past frame (for example, the previous frame) to block each block of the frame that is the current encoding target (processing target). The vector quantization distortion e corresponding to this vector quantization is calculated as shown in Expression (6). Calculated using the vector quantization distortion calculated from the vector quantization table generated from the block in the current encoding target frame and the vector quantization table generated from the past frame. Compare vector quantization distortion, select (adopt) the vector quantization table with the smaller vector quantization distortion, and encode the vector quantization result using the selected vector quantization table It is encoded by a method (for example, Huffman encoding) and output.

  In the decoding device 200 according to the second embodiment, when the input encoded signal is a vector quantized signal using a vector quantization table generated in a past frame, the inverse vector quantization unit 6 performs decoding. The signal decoded by the unit 5 is subjected to inverse vector quantization using the vector quantization table generated in the past frame.

  On the other hand, when the encoded signal input to the decoding device 200 is a signal that has been vector quantized using a vector quantization table generated from a block in the same frame, the inverse vector quantization unit 6 includes a decoding unit. A vector quantization table is generated from the frequency transform coefficients for one block (for example, the first block on the time axis) decoded in step 5, and the generated vector quantization table is used to generate another vector quantization table. The block is subjected to inverse vector quantization for each block, and the signal after inverse vector quantization is output to the frequency inverse transform unit 7.

Next, the operation in the second embodiment will be described.
First, an encoding process executed in the encoding device 100 according to the second embodiment will be described with reference to the flowchart of FIG.

  First, the input signal is divided into frames (step S10), and the input signal is subjected to frequency conversion in units of blocks constituting one frame (step S11). Next, the frequency transform coefficient for one block in one frame (for example, the first block on the time axis) is encoded by a predetermined encoding method (for example, Huffman encoding) (step S12), and the encoding is performed. A vector quantization table is generated from the frequency transform coefficients of the blocks (step S13).

  Next, using the vector quantization table generated in step S13, vector quantization is performed on the frequency transform coefficients of other blocks in the same frame (blocks other than the block encoded in step S12) (step S14). ), Vector quantization distortion is calculated from the result of vector quantization in step S14 as shown in equation (6) (step S15).

  Next, using the vector quantization table generated in the past frame (for example, the previous frame), vector quantization is performed on the frequency transform coefficient of each block of the current frame to be encoded. Then, the vector quantization distortion is calculated from the vector quantization result in step S16 as shown in equation (6) (step S17).

  Next, the vector quantization distortion calculated in step S15 is compared with the vector quantization distortion calculated in step S17, and the vector quantization table having the smaller vector quantization distortion value is selected (step S18). The vector quantization result using the selected vector quantization table is encoded by a predetermined encoding method (for example, Huffman encoding) (step S19), and the encoding process ends.

  Next, the decoding process executed in the decoding device 200 according to the second embodiment will be described with reference to the flowchart of FIG.

  First, based on the encoded signal input to the decoding apparatus 200, it is determined whether or not the vector quantization table generated in the past frame has been used in the vector quantization on the encoding side (step T10). . If it is determined in step T10 that the vector quantization table generated from the block in the current decoding target (processing target) frame is used (step T10; NO), the encoded signal for one block in the frame Is decoded (step T11), and a vector quantization table is generated from the frequency transform coefficients for one block obtained by the decoding (step T12).

  Next, the coded signal of the other block in the current decoding target frame is decoded, and the frequency transform coefficient of each block obtained by decoding is used for each block using the vector quantization table generated in step T12. Is subjected to inverse vector quantization (step T13). Next, frequency inverse transform is performed on the signals obtained in steps T11 and T13 (step T14), and the decoding process ends.

  If it is determined in step T10 that the vector quantization table generated in the past frame is used (step T10; YES), the vector quantization table generated in the past frame is prepared (step T15). The encoded signal of each block of the current decoding target frame is decoded, and the inverse vector quanta for each block using the vector quantization table prepared in step T15 is obtained for the frequency transform coefficient of each block obtained by decoding. (Step T13). Next, frequency inverse transformation is performed on the signal obtained in step T13 (step T14), and the decoding process ends. This decoding process is performed for each frame, and the decoded frames are combined and output as a reproduction signal.

  As described above, according to the second embodiment of the present invention, it is possible to reduce the overall code amount by performing vector quantization using the vector quantization table generated in the past frame.

The block diagram (a) which shows the structure of the encoding apparatus which concerns on Embodiment 1 of this invention, and the block diagram (b) which shows the structure of the decoding apparatus which concerns on Embodiment 1 of this invention. The figure which shows an example of a vector quantization table. 3 is a flowchart showing an encoding process executed in the encoding apparatus according to the first embodiment. 5 is a flowchart showing a decoding process executed in the decoding apparatus according to the first embodiment. 9 is a flowchart showing an encoding process executed in the encoding apparatus according to the second embodiment. 9 is a flowchart illustrating a decoding process executed in the decoding device according to the second embodiment.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Frame conversion part 2 Frequency conversion part 3 Encoding part 4 Vector quantization part 5 Decoding part 6 Inverse vector quantization part 7 Frequency reverse conversion part 8 Frame synthetic | combination part 31 Vector quantization table 100 Encoding apparatus 200 Decoding apparatus

Claims (6)

Framing means for dividing the input signal into frames;
Frequency conversion means for performing frequency conversion on the input signal for each block unit constituting one frame;
Encoding means for encoding a frequency conversion coefficient for a predetermined block in one frame;
Generating means for generating a vector quantization table based on a frequency conversion coefficient for one block that is the target of encoding by the encoding means;
Vector quantization means for performing vector quantization on frequency transform coefficients of other blocks in the same frame that are not encoded by the encoding means, using the vector quantization table generated by the generating means ,
An encoding device comprising: Calculating means for calculating a frequency distribution of frequency conversion coefficient values for each block unit constituting one frame;
Selecting means for selecting a frequency transform coefficient for one block used for generating the vector quantization table based on the frequency distribution calculated by the calculating means;
The encoding apparatus according to claim 1, wherein the encoding unit uses the frequency conversion coefficient for one block selected by the selection unit as the frequency conversion coefficient for the predetermined one block . The encoding apparatus according to claim 1, wherein the encoding unit uses a frequency conversion coefficient of a first block of one frame as the frequency conversion coefficient for the predetermined one block . A decoding device for decoding an input signal encoded by the encoding device according to claim 1,
Decoding means for decoding a predetermined one block of frequency transform coefficients in one frame encoded by the encoding means of the encoding device;
Generating means for generating a vector quantization table based on the frequency transform coefficients for one block decoded by the decoding means ;
A signal obtained by performing vector quantization on the frequency transform coefficient of another block in the same frame by the vector quantization means of the encoding device is inverted using the vector quantization table generated by the generation means. Inverse vector quantization means for performing quantization;
Frequency inverse transform means for performing frequency inverse transform on the frequency transform coefficient obtained by the inverse vector quantization;
A decoding apparatus comprising:   A framing step for dividing the input signal into frames;
  A frequency conversion step of performing frequency conversion on the input signal for each block unit constituting one frame;
  An encoding step for encoding a frequency transform coefficient for a predetermined block in one frame;
  A generation step of generating a vector quantization table based on a frequency transform coefficient for one block that is an encoding target in the encoding step;
  A vector quantization step of performing vector quantization on the frequency transform coefficients of other blocks in the same frame that are not encoded in the encoding step, using the vector quantization table generated in the generation step; ,
An encoding method comprising: A decoding method for decoding an input signal encoded by the encoding method according to claim 5, comprising:
  A decoding step of decoding a predetermined one block of frequency transform coefficients in one frame encoded by the encoding step of the encoding method;
  A generating step for generating a vector quantization table based on the frequency transform coefficients for one block decoded by the decoding step;
  A signal obtained by performing vector quantization on the frequency transform coefficient of another block in the same frame by the vector quantization step of the encoding method is inverted using the vector quantization table generated by the generation step. An inverse vector quantization step for performing quantization;
  A frequency inverse transform step for performing frequency inverse transform on the frequency transform coefficient obtained by the inverse vector quantization;
A decoding method characterized by comprising:

Images