JP2009081782A - Lossless compression coding method, apparatus thereof, and lossless compression expanding method, and apparatus thereof - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To improve compression efficiency by solving a problem wherein, when the level of an input sample is small, its compression efficiency deteriorates owing to Huffman coding processing to data subjected to common redundance portion removal in order to convert the data into Huffman coding data. <P>SOLUTION: The methods and appratuses includes: a common redundancy portion removing means 12 to remove a redundancy portion common to each sample constituting a block; an high-order portion/low-order portion separating means 13 which separates data so that data not subjected to Huffman processing but subjected to common redundancy portion removal is superposed on a bit stream when the bit length of the data from which the common redundancy portion is removed is equal or less than the minimum bit length of data subjected to Huffman coding; and a multiplexer 16 which prevents supplementary information HI specifing a Huffman coding algorithm from being superposed on the bit stream when the bit length of the data subjected to common redundancy portion removal is equal to or less than the minimum bit length of data subjected to Huffman coding. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to encoding and decoding of lossless (reversible) compression for transmitting and recording a digital audio signal.

  There are various methods for compressing a digital audio signal, and these compression methods are roughly classified into two methods. The first is transmission recording by reducing the amount of information by compressing only redundant components that depend on autocorrelation and cross-correlation of sample data in which digital audio signals are arranged in time series, or correlation between channels of multiple channels. This encoding method is called a lossless compression method. The second is a code that uses human auditory characteristics to obtain high efficiency, allows the omission of insensitive components that are difficult to perceive (ie, degrades), and greatly reduces the amount of information transmitted and recorded. And is referred to as a lossy compression method. The latter is generally used together with the former and is frequently used in communication means and recording media having a low transmission recording bit rate. However, when a high quality sound quality is required, the former lossless compression method without sound quality deterioration must be used. In the following, the description will be limited to lossless compression.

  As an example of the prior art of the lossless compression method, there is a “lossless compression encoding method and apparatus and lossless compression decoding method and apparatus” described in Patent Document 1. The purpose of this prior art is a 96 kHz, 24 bit, 6 ch mode, which is a huge amount of information, and 74 minutes of digital audio data with high quality sound without deterioration and limited to, for example, a 4.7 GB DVD The present invention provides a lossless compression encoding / decoding method and apparatus for reducing the amount of information of original data for recording on a recording medium having a capacity.

  Therefore, in the above-described lossless compression of the prior art, the correlation component between samples is removed in units of a block consisting of a predetermined number of samples obtained by dividing a time-series input sample represented by a predetermined bit length every predetermined time. At the same time, auxiliary information specifying the correlation removal characteristic is obtained. The redundant run length is obtained for the sample after removing the correlation component, and the minimum value of the redundant run length in the block, that is, the run length of the redundant portion common to each sample in the block (hereinafter referred to as a common redundant portion) is detected. Then, the common redundant part is removed from each sample, and the run length information of the common redundant part is obtained. The data from which the common redundant part is removed is Huffman-encoded to obtain Huffman-encoded data, and auxiliary information for specifying the Huffman encoding algorithm is obtained. The run length information of the common redundant portion obtained in this way, the Huffman encoded data for each sample, and auxiliary information specifying the correlation removal characteristics and the Huffman encoding algorithm are superimposed on the bitstream.

  In this way, in the lossless compression described above, the data after removing the correlation between the samples of the input samples and removing the common redundant part is further effectively compressed using the data distribution bias, Compress to the amount of information close to the information entropy of the input sample.

  Further, in removing the correlation component between samples, the difference between the data predicted by the predictor or the prediction filter and the actual data, that is, the prediction residual is extracted. For example, these predictors or prediction filters are composed of integer arithmetic circuits having an integer range of input samples or more.

  In addition, a plurality of correlation removal characteristics are provided, and a correlation removal characteristic that provides the maximum compression efficiency is selected and used, and auxiliary information that specifies the selected correlation removal characteristic is superimposed on the bitstream.

  Also, in Huffman coding, Huffman coding is performed only on the upper part data of a predetermined bit length from the data from which the common redundant part is removed, and if there is remaining lower part data, it is directly superimposed on the bit stream.

  In addition, a plurality of Huffman encoding algorithms are provided, and the Huffman encoding algorithm that provides the maximum compression efficiency is selected and used, and auxiliary information that specifies the selected Huffman encoding algorithm is superimposed on the bitstream.

  Further, in the above-described lossless compression decoding of the prior art, a bit stream generated by the above-described lossless compression encoding is input, auxiliary information specifying a Huffman encoding algorithm is extracted, and Huffman decoding is performed. Further, the run length information of the common redundant part in the block is extracted, and the common redundant part in the block is added to the data after the Huffman decoding. Further, auxiliary information specifying the correlation removal characteristic is extracted, and a process of restoring the correlation component between samples is performed. In this way, in the above-described lossless compression decoding of the prior art, the original digital audio data is completely restored from the input bit stream without deterioration and output.

  Also, in Huffman decoding, Huffman decoding is performed only on the upper part data by dividing the upper part data corresponding to the predetermined Huffman codeword into lower part data having a predetermined bit length based on the run length information of the common redundant part. Apply.

  Also, in Huffman decoding, auxiliary information specifying a Huffman encoding algorithm is extracted, and one of a plurality of Huffman encoding tables provided based on this auxiliary information is selected to match a predetermined Huffman codeword Huffman decoding is performed on the part to be restored to restore the upper part data.

  Further, in the process of restoring the correlation component, the sum of the data predicted by the predictor or the prediction filter and the actual data is extracted. For example, these predictors or prediction filters are composed of integer arithmetic circuits having an integer range of output samples or more.

  FIG. 6 is a block diagram of the above-described lossless compression encoding apparatus 1 of the prior art. In FIG. 6, the correlation component removing unit 11 uses time-series input samples IN (i) (i is 0 <= i <N) with a block made up of a predetermined number N of samples divided every predetermined time as a unit. Data D1 (j, i) from which correlation components between (integer) samples are removed is output. The correlation component removal unit 11 has a plurality of correlation removal characteristics, and j represents an index for designating one of the plurality of correlation removal characteristics.

  Further, the common redundant part removing unit 12 calculates a redundant run length representing a length in which the same bit value continues in the bit string for each sample of the data D1 (j, i) after the correlation component is removed by the correlation component removing unit 11. At the same time, the minimum value of the redundant run length in the block, that is, the run length of the common redundant portion of the block is detected, and data D2 (j, i) from which the common redundant portion is removed is output. Further, the common redundant part removing unit 12 outputs the run length information CRB (j) of the common redundant part.

  Further, the upper part / lower part separating means 13 performs the upper part data U1 (j, i) having a predetermined bit length subjected to Huffman coding and the remaining lower part from the data D2 (j, i) from which the common redundant part is removed. Each is divided into partial data L1 (j, i) and output.

  The Huffman encoding means 14 performs Huffman encoding on the separated higher-order data U1 (j, i) and outputs Huffman encoded data H1 (k, j, i). Here, the Huffman encoding means 14 includes a plurality of Huffman encoding algorithms, and k represents an index for designating one of the plurality of Huffman encoding algorithms.

  Further, the maximum efficiency data output means 15 has a correlation removal characteristic and a Huffman so that the maximum compression efficiency is obtained from the combination of the lower order data L1 (j, i) and the Huffman encoded data H1 (k, j, i). Any one of the encoding algorithms is selected, and lower order data L2 (i) and Huffman encoded data H2 (i) to be superimposed on the bit stream are output. Further, the index j specifying the selected correlation removal characteristic is output as auxiliary information DI, and the index k specifying the selected Huffman coding algorithm is output as auxiliary information HI.

  The multiplexer 16 includes the run length information CRB (DI) of the common redundant part in the block, the auxiliary information DI specifying the correlation removal characteristic, the auxiliary information HI specifying the Huffman coding algorithm, and the lower part data L2 (i). And Huffman encoded data H2 (i) are superimposed on the bit stream ST and output.

  FIG. 7 is an internal block diagram showing an example of the correlation component removing unit 11 of FIG. In FIG. 7, the correlation component removing unit 11 includes two delay units and two arithmetic units. The input sample IN (i) of the time series input is one sample period before the delay by the delay unit 111. A difference from the input sample IN (i−1) is obtained by the calculator 112, and the primary difference value obtained in this way is output as data D1 (1, i) from which the correlation component is removed. Further, the difference between the data D1 (1, i) represented by the primary difference value and the data D1 (1, i-1) of the primary difference value before the one sample period delayed by the delay unit 113. Is obtained by the computing unit 114, and the secondary difference value thus obtained is output as data D1 (2, i) from which the correlation component is removed. However, in FIG. 7, the input sample IN (i) can be selected when it is preferable to select the input sample IN (i) from which the correlation component is not removed in order to obtain the maximum compression efficiency. The input sample IN (i) is output as D1 (0, i).

  In FIG. 7, as a method for removing the correlation component of the input sample for the sake of simplicity, the primary difference value and the secondary difference value are obtained and output as data from which the correlation component is removed. The correlation component may be removed using a detector or a prediction filter. Furthermore, you may make it have three or more correlation removal characteristics.

  FIG. 8 is an internal block diagram of the common redundant part removing means 12 of FIG. In FIG. 8, first, data D1 (j, i) for the number of samples N of the block is input to the delay buffer 121 and held. The redundant run length detection unit 122 sequentially obtains redundant run lengths representing the length of the same bit value in the bit string for each sample of D1 (j, i). The common redundant run length detection unit 123 detects the minimum redundant run length obtained for each sample, that is, the run length of the redundant portion common to each sample of the block, and outputs the run length information CRB (j) of the common redundant portion. To do. The common redundant part removing unit 124 outputs data D2 (j, i) obtained by removing the common redundant part of D1 (j, i) based on the run length information CRB (j) of the common redundant part.

  In the following, for simplicity of explanation, the number of samples N in the block is 6 and the bit length WL per input sample is 16 bits. Further, the bit length IB of the index IDX having a predetermined bit length in the Huffman coding table used in the Huffman coding means is 6 bits.

  FIG. 9 shows data D1 (j, i) after removing the correlation component, redundant run length RB (j, i) detected for each sample, data D2 (j, i) from which the common redundant part is removed, common It is explanatory drawing showing the run length information CRB (j) of a redundant part. The redundant run length detection unit 122 in FIG. 8 detects the redundant run length RB (j, i) for each sample of the data D1 (j, i). In FIG. 9, redundant run lengths RB (j, i) = {7, 6, 8, 12, 9, 10} are detected.

  The common redundant run length detection unit 123 uses the minimum redundant RB (j, 1) = 6 among the redundant run lengths RB (j, i) as the common redundant run length information CRB (j) indicating the run length of the common redundant part of the block. = 6 is output. The common redundant part removing unit 124 outputs 10-bit data D2 (j, i) obtained by removing the common redundant part from the data D1 (j, i) based on the common redundant run length information CRB (j) = 6. To do.

  FIG. 10 shows the data D2 (j, i) from which the common redundant part is inputted, the separated upper part data U1 (j, i) and the lower part in the upper part / lower part separating means 13 of FIG. It is explanatory drawing showing data L1 (j, i). In FIG. 10, the upper part / lower part separating means 13 converts the bit length IB of the index IDX of the Huffman coding table from the 10-bit data D2 (j, i) from which the common redundant part is removed, that is, the upper 6 bits to the upper part. Data U1 (j, i) is output and the remaining lower 4 bits are output as lower data L1 (j, i).

  FIG. 11 is a table showing an example of a Huffman encoding table used in the Huffman encoding algorithm applied to the higher order data U1 (j, i) input by the Huffman encoding means 14 of FIG. The Huffman code word HCW (IDX) having the corresponding code length HLN (IDX) is selected with the higher order data U1 (j, i) as the index IDX, and is output as Huffman encoded data H1 (k, j, i).

  For example, in the case of U1 (j, i) = [000000] (inside [] represents a bit string pattern), a Huffman codeword HCW (0) = [0000] having a code length HLN (0) of 4 bits. When U1 (j, i) is [110111], the Huffman code word HCW (55) = [1111110111] having a code length HLN (55) of 10 bits is output. In general, in the Huffman coding table used in the Huffman coding algorithm, a short code length Huffman code word is used for an index with a high appearance frequency in an index appearance frequency distribution, and a long code length is used for an index with a low appearance frequency. A variable-length Huffman codeword is assigned in the form of a Huffman codeword. By doing so, a short code length is assigned to an index having a high appearance frequency, and the total amount of information can be reduced. In addition, by providing a plurality of Huffman encoding algorithms having different index appearance frequency distributions, a Huffman encoding algorithm that provides the maximum compression efficiency is selected and used.

  12 shows the Huffman coding when the upper part data U1 (j, i) output from the upper part / lower part separating means 13 of FIG. 6 is subjected to Huffman coding using the Huffman coding table of FIG. It is explanatory drawing showing data H1 (1, j, i). Using the higher order data U1 (j, i) as an index IDX, the corresponding Huffman code word HCW (IDX) in the Huffman coding table of FIG. 11 is converted and output as Huffman coded data H1 (1, j, i). In FIG. 12, the total number of bits of the 6-bit upper part data U1 (j, i) is 36 bits (= 6 bits × 6 samples), whereas the output Huffman encoded data H1 (1, This represents that the total number of bits of j, i) is 33 bits.

  However, depending on the combination of the upper part data U1 (j, i), a lot of data that appears to have a low appearance frequency appears, and as a result, a Huffman codeword HCW (IDX) having a code length HLN (IDX) longer than the original 6 bits ), And the total number of bits of the Huffman encoded data H1 (k, j, i) may be larger than the total number of bits of the higher order data U1 (j, i). As described above, in order to maximize the compression efficiency, when it is preferable to transmit the high order data U1 (j, i), the high order data U1 (j, i) is converted to Huffman so that it can be selected. Output as encoded data H1 (0, j, i).

  FIG. 13 shows the run length of the common redundant part in the block detected by the common redundant run length detection part 123 of FIG. 8 from the bit length WL per sample of the data D1 (j, i) to the index of the Huffman coding table. It is explanatory drawing showing the removal of a common redundant part in the case where it is more than (WL-IB) which deducted bit length IB. In FIG. 13, when the run length of the common redundant portion in the block is (WL-IB) or more, the run length of the common redundant portion in the block is set to (WL-HB), and the common redundant portion is removed. That is, the common redundant run length information CRB (j) is set to (WL-IB), and the data D2 (j, i) from which the common redundant portion is removed becomes the lower IB bit portion of the data D1 (j, i). When the common redundant part is removed in this way, the upper part data U1 (j, i) separated by the upper part / lower part separating means 13 from the data D2 (j, i) from which the common redundant part is removed is shared. It is the same as the data D2 (j, i) from which the redundant part is removed, and the lower part data L1 (j, i) is not output.

  FIG. 14 is an explanatory diagram showing the Huffman encoded data H1 (k, j, i) when the Huffman encoding is applied to the higher order data U1 (j, i) shown in FIG. In FIG. 14, the total number of bits of the 6-bit upper part data U1 (j, i) is 36 bits (= 6 bits × 6 samples), whereas the output Huffman encoded data H1 (k, The total number of bits j, i) is reduced to 24 bits.

  FIG. 15 is a configuration diagram of the above-described lossless compression decoding apparatus 2 of the prior art. In FIG. 15, the demultiplexer 21 includes common redundant run-length information CRB2 indicating the run length of the common redundant portion in the block from the input bit stream ST, auxiliary information DI specifying the correlation removal characteristic, and a Huffman coding algorithm. Auxiliary information HI for designating, subordinate data L3 (i) and Huffman encoded data H3 (i) are extracted.

  The Huffman decoding means 22 includes Huffman encoded data H3 (i) represented by a bit string extracted by the code length HLN (IDX) of the Huffman encoding table indicated by the auxiliary information HI designating the Huffman encoding algorithm, and Huffman encoding. A match with the Huffman code word HCW (IDX) in the table is detected, and the index IDX corresponding to the matching Huffman code word HCW (IDX) is output as the upper data U2 (i). Here, the output higher order data U2 (i) has the same predetermined bit length as the higher order data U1 (j, i) input to the Huffman encoding means 14 of FIG.

  The upper part / lower part combining means 23 outputs data D3 (i) obtained by combining the upper part data U2 (i) and the lower part data L3 (i). The common redundant part adding means 24 outputs data D4 (i) obtained by adding the common redundant part to the data D3 (i) based on the run length information CRB2 of the common redundant part in the block. The data D4 (i) with the common redundant part added has the same bit length as the data D1 (j, i) input to the common redundant part removing means 12 in FIG. The correlation component restoration unit 25 performs the inverse transformation process of the correlation component removal unit 11 of FIG. 6 on the data D4 (i) to which the common redundant portion is added based on the auxiliary information DI that specifies the correlation removal characteristic, and outputs the output sample. OUT (i) is restored and output.

  FIG. 16 shows Huffman encoded data H3 (i) and lower part data L3 (i) extracted from the bitstream ST, upper part data U2 (i) restored from Huffman encoded data H3 (i), and upper part It is explanatory drawing showing data D3 (i) which combined partial data U2 (i) and lower order data L3 (i).

  Huffman decoding means 22 in FIG. 15 performs Huffman encoded data H3 (i) represented by a bit string extracted with code length HLN (IDX) based on the Huffman encoding table in FIG. 11 and a Huffman codeword in the Huffman encoding table. Match detection with HCW (IDX) is performed, and the index IDX corresponding to the matching Huffman codeword HCW (IDX) is output as the higher-order data U2 (i). The upper part data U2 (i) restored in this way and the lower part data L3 (i) extracted from the bit stream ST are combined by the upper part / lower part combining means 23 and output as data D3 (i). . The bit length of the data D3 (i) is (WL-CRB2) obtained by subtracting the run length CRB2 of the common redundant portion of the block from the predetermined bit length WL per sample of the output samples.

  As described above, in the lossless compression coding of the prior art, the correlation component removing unit removes the correlation of the input samples, the common redundant part is removed, and the Huffman compression unit uses the bias of the appearance frequency distribution of the data. It converts to Huffman encoded data, further compresses the data, adds auxiliary information, and generates a bitstream with a smaller amount of information than the input samples.

Further, in the lossless compression decoding of the prior art, the Huffman decoding is performed on the Huffman encoded data extracted from the bitstream by the Huffman decoding means, a common redundant part is added, and the correlation is restored so that the degradation is completely eliminated. Output missing output samples. As a result, it is possible to compress the amount of information close to the information entropy of the input sample, and it is possible to record / reproduce data on a recording medium having a limited storage capacity for a longer time.
Japanese Patent No. 3723740

  However, in the lossless compression coding of the prior art, when a small level sample is input, the common redundant part of the block is removed so that the data of the bit length IB of the index IDX of the predetermined bit length of the Huffman coding table remains. In addition, in order to perform Huffman coding using the data from which the common redundant part is removed as the upper part data, at least the Huffman coded data having the minimum bit length of the Huffman codeword and auxiliary information for designating the Huffman coding algorithm are bits. There is a problem in that the compression efficiency is lowered because the stream is superimposed on the stream.

  The lossless compression encoding method of the present invention is expressed by a predetermined bit length, and a block composed of a predetermined number of samples divided every predetermined time from sample data input side by side in time series. A common component common to each sample having a bit length corresponding to the minimum redundant run length of each sample in the block from the correlation component removing step for removing the correlation component between samples and the data after removing the correlation component A common redundant part removing step for removing the redundant part, and if the bit length of the data from which the common redundant part is removed is equal to or smaller than the minimum bit length of the data obtained by Huffman coding, the Huffman coding is not performed. When the bit length of the data from which the common redundant part is removed is larger than the minimum bit length of the data obtained by the Huffman coding, the common Huffman encoding step for obtaining Huffman encoded data by Huffman encoding the data from which the long part has been removed, the run length information of the common redundant part, the data from which the common redundant part is removed for each sample, or the Huffman for each sample A superimposing step of superimposing the encoded data on the bit stream.

  Further, the data from which the common redundant part is removed is further provided with an upper part / lower part separating step for separating the data from which the common redundant part is removed into upper part data having a predetermined bit length and the remaining lower part data. Is larger than the minimum bit length of data obtained by Huffman coding, in the Huffman coding step, the higher order data is Huffman coded to obtain Huffman coded data, and in the superposition step, The lower part data is superimposed on the bit stream.

  Further, the Huffman encoding step has a plurality of Huffman encoding algorithms, and further, the bit length of the data after removing the common redundant portion is the data obtained by Huffman encoding of all the Huffman encoding algorithms. A maximum efficiency algorithm selection step of selecting a Huffman encoding algorithm that maximizes the compression efficiency in the block from the plurality of Huffman encoding algorithms when the bit length is greater than the minimum bit length, Furthermore, a parameter indicating the selected Huffman encoding algorithm is superimposed on the bit stream.

  In the superimposing step, when the bit length of the data from which the common redundant portion is removed is equal to or less than the minimum bit length of data obtained by Huffman coding of all the Huffman coding algorithms, the parameter is added to the bit stream. It does not overlap.

  Further, the lossless compression decoding method of the present invention is a lossless compression decoding using a bit stream in which run length information of a common redundant portion and data obtained by removing the common redundant portion for each sample or Huffman encoded data for each sample are superimposed. An extraction step of extracting run length information of the common redundant portion and data obtained by removing the common redundant portion for each sample or Huffman encoded data for each sample from the bitstream, and the extracted sample A Huffman decoding step of performing Huffman decoding on each Huffman encoded data and outputting Huffman decoded data, and a bit length obtained by subtracting a bit length of the run length information of the extracted common redundant portion from the predetermined bit length of the sample The minimum bit of data obtained by the Huffman coding. When larger than the length, add a common redundant part based on the run length information of the extracted common redundant part to the Huffman decoded data Huffman decoded in the Huffman decoding step, the extracted from the predetermined bit length of the sample When the bit length obtained by subtracting the bit length of the run length information of the common redundant portion is equal to or smaller than the minimum bit length of the data obtained by the Huffman coding, the extracted data is obtained by removing the common redundant portion for each of the extracted samples. A common redundant portion adding step for adding a common redundant portion based on the run length information of the common redundant portion, and a correlation for restoring the original sample by adding a correlation component between the samples to the data to which the common redundant portion is added A component restoration step.

  Further, the bit stream is further superimposed with lower part data obtained by removing upper part data having a bit length corresponding to the codeword length of the Huffman coding from data obtained by removing the common redundant part for each sample, In the extracting step, the low order part data is extracted, and further includes a high order part / low order part combining step of combining the Huffman decoded data as a high order part and combining the extracted low order part data as a low order part. The Huffman decoding step when the bit length obtained by subtracting the extracted bit length of the run length information of the common redundant portion from the predetermined bit length of the sample is greater than the post-code length of the Huffman coding In place of the Huffman-decoded data that was Huffman-decoded in Step 1, the higher-order / lower-order combination step It is characterized in adding a common redundant portion on the basis of the said extracted common redundant portion runlength information data.

  Further, Huffman encoding algorithm information indicating one of a plurality of Huffman encoding algorithms is superimposed on the bitstream, and in the extraction step, the Huffman encoding algorithm information is extracted, and in the Huffman decoding step, , Huffman decoding is performed using a Huffman encoding algorithm based on the extracted Huffman encoding algorithm information.

  With the above configuration, the lossless compression encoding method of the present invention prevents Huffman encoding when the bit length of the data from which the common redundant portion is removed is equal to or less than the minimum bit length of the data obtained by Huffman encoding. As a result, the total number of bits of data to be superimposed on the bit stream can be reduced. At this time, the total number of bits can be further reduced by not superimposing the parameter indicating the Huffman coding algorithm on the bit stream.

  Also, the lossless compression decoding method of the present invention can restore the original sample from the bitstream whose total number of bits has been reduced by the lossless compression encoding method.

  The best mode for carrying out the present invention will be described below with reference to the drawings.

(Embodiment 1)
FIG. 1 is a configuration diagram of a lossless compression coding apparatus 1 according to Embodiment 1 of the present invention. In FIG. 1, the correlation component removing unit 11 uses time-series input samples IN (i) (i is 0 <= i <N) with a block composed of a predetermined number N of samples divided every predetermined time as a unit. Data D1 (j, i) from which correlation components between (integer) samples are removed is output. The correlation component removal unit 11 includes a plurality of correlation removal characteristics, and j represents an index that designates one of the plurality of correlation removal characteristics.

  Further, the common redundant part removing unit 12 calculates a redundant run length representing a length in which the same bit value continues in the bit string for each sample of the data D1 (j, i) after the correlation component is removed by the correlation component removing unit 11. At the same time, the minimum value of the redundant run length in the block, that is, the run length of the common redundant portion in the block is detected, and data D2 (j, i) from which the common redundant portion is removed is output. Further, the common redundant part removing unit 12 outputs run length information CRB (j) indicating the bit length of the common redundant part.

  Further, the upper part / lower part separating means 13 performs the upper part data U1 (j, i) having a predetermined bit length to be subjected to Huffman coding from the data D2 (j, i) from which the common redundant part is removed, and the remaining Each is divided into lower order data L1 (j, i) and output. The Huffman encoding means 14 performs Huffman encoding on the separated higher-order data U1 (j, i) and outputs Huffman encoded data H1 (k, j, i). Here, the Huffman encoding means 14 includes a plurality of Huffman encoding algorithms, and k represents an index for designating one of the plurality of Huffman encoding algorithms.

  Further, the maximum efficiency data output means 15 has a correlation removal characteristic and a Huffman so that the maximum compression efficiency is obtained from the combination of the lower order data L1 (j, i) and the Huffman encoded data H1 (k, j, i). Any one of the encoding algorithms is selected, and lower order data L2 (i) and Huffman encoded data H2 (j) to be superimposed on the bit stream are output. Further, an index j specifying the selected correlation removal characteristic is output as auxiliary information DI. In addition, an index k specifying the selected Huffman coding algorithm is output as auxiliary information HI.

  The multiplexer 16 includes the run length information CRB (DI) of the common redundant part in the block, the auxiliary information DI specifying the correlation removal characteristic, the auxiliary information HI specifying the Huffman coding algorithm, and the lower part data L2 (i). And Huffman encoded data H2 (i) are superimposed on the bit stream ST and output.

  Here, in FIG. 1, the correlation component removing unit 11 is the same as the above-described lossless compression coding apparatus of the prior art.

  However, the run length of the common redundant portion in the block detected by the common redundant run length detection unit 123 of FIG. 8 is calculated from the bit length WL per sample of the data D1 (j, i) from the Huffman codeword of the Huffman coding table. The common redundant portion removal method when the minimum value HB of the code length HLN is equal to or greater than (WL-HB) is different from the common redundant portion removal method in the lossless compression coding apparatus of the prior art described above. The processing contents of the common redundant part removing means 12, the upper part / lower part separating means 13, the Huffman coding means 14 and the maximum efficiency data output means 15 are different. However, the processing in the case where the run length of the common redundant portion is smaller than (WL-HB) is the same as that of the above-described lossless compression encoding device of the prior art.

  FIG. 2 shows that the run length of the common redundant portion in the block detected by the common redundant run length detection unit 123 is Huffman-coded from the bit length WL (16 bits in this case) per sample of the data D1 (j, i). It is explanatory drawing showing the removal of a common redundant part in the case of (WL-HB) (here 12 bits) more than the minimum value HB (here 4 bits) of code length HLN of the Huffman codeword of a table. In the lossless compression coding apparatus according to the present invention, as shown in FIG. 2, even when the run length of the common redundant portion is (WL-HB) or more, the common redundant portion is detected with the detected bit length of the run length of the common redundant portion. Remove. That is, the data D2 (j, i) from which the common redundant portion is removed becomes a lower (WL-CRB (j)) bit portion of the data D1 (j, i). When the common redundant part is removed in this way, the upper part / lower part separating means 13 does not output the upper part data U1 (j, i), but the data D2 (j, i) is output as the lower part data L1 (j, i).

  As a result, the total number of bits of data to be superimposed on the bit stream becomes 18 bits (= 3 bits × 6 samples), and the run length of the common redundant portion is (WL-HB) in the above-described lossless compression coding of the prior art. ), The data D2 (j, i) after removing the common redundant portion is used as Huffman coding data as the upper portion data U1 (j, i). The amount of information can be reduced compared to bits. According to the first embodiment of the present invention, when the run length of the common redundant part in the block is (WL-HB) or more, the auxiliary information HI specifying the Huffman coding algorithm is not superimposed on the bitstream. By doing so, the amount of information can be further reduced.

  Further, as shown in FIG. 2, only the lower part data L1 (j, i) is output from the data D2 (j, i) from which the common redundant part is removed, and the upper part data U1 (j, i) is not output. Inputs the run length information CRB (j) of the common redundant portion to the Huffman encoding unit 14 so that the Huffman encoding unit 14 does not perform a series of processes related to the Huffman encoding. In this way, processing related to Huffman coding can be reduced. In such a case, since the Huffman encoded data H1 (k, j, i) is not output from the Huffman encoding means 14, the Huffman encoded data H1 (k, j, In order not to use i), the run-length information CRB (j) of the common redundant portion is inputted to the maximum efficiency data output means 15. Further, the multiplexer 16 prevents the auxiliary information HI specifying the Huffman coding algorithm from being superimposed on the bitstream based on the input run length information CRB (j) of the common redundant portion.

  Here, in the case where the bit length WL per sample of the data D1 (j, i) is 16 bits and the Huffman coding table is as shown in FIG. 11, the effect of reducing the amount of information according to the first embodiment of the present invention. Will be described with reference to FIG.

  FIG. 3 shows the presence / absence of the Huffman encoding process and the bit length of the higher order data U (j, i) in each case where the bit length of the common redundant portion is 9 to 15 bits in the case of the conventional technique and the first embodiment. , The bit length of the lower part data L (j, i), the bit length of the Huffman codeword, and the total bit length of the lower part data and the Huffman codeword. From FIG. 3, when the common redundant part bit length is 12 bits or more, that is, (WL-HB) or more, the total bit length of the first embodiment becomes smaller than the total bit length of the prior art, and the amount of information is reduced. You can see that it has been reduced.

  Also, when the bit length of the common redundant part is 11 bits, if the Huffman coding is not performed as in the case where the bit length of the common redundant part is 12 bits or more, the bit length of the lower order data is 5 It can be seen that the total bit length is not necessarily smaller than that of the prior art because it is a bit.

  By the way, since the bit length CRB (j) of the common redundant portion changes to (WL-HB) or less when the index j specifying the correlation removal characteristic is different, the maximum efficiency data output means 15, data based on different processes are mixedly input, and it is only necessary to detect the data having the maximum efficiency.

  FIG. 4 is a configuration diagram of the lossless compression decoding apparatus 2 according to Embodiment 1 of the present invention. In FIG. 4, the demultiplexer 21 receives common redundant run-length information CRB2 indicating the run length of the common redundant portion in the block from the input bit stream ST, auxiliary information DI specifying the correlation removal characteristic, and a Huffman encoding algorithm. The auxiliary information HI to be designated, the lower part data L3 (i), and the Huffman encoded data H3 (i) are extracted.

  The Huffman decoding means 22 includes Huffman encoded data H3 (i) represented by a bit string extracted by the code length HLN (IDX) of the Huffman encoding table indicated by the auxiliary information HI designating the Huffman encoding algorithm, and Huffman encoding. A match with the Huffman code word HCW (IDX) in the table is detected, and the index IDX corresponding to the matching Huffman code word HCW (IDX) is output as the upper data U2 (i). Here, the output higher order data U2 (i) has the same predetermined bit length as the higher order data U1 (j, i) input to the Huffman encoding means 14 of FIG. However, based on the run length information CRB2 of the common redundant part, when the run length of the common redundant part is (WL-HB) or more, the Huffman encoded data is not superimposed on the bit stream ST. The processing related to data decoding is not executed, and the higher order data U2 (i) is not output.

  The upper part / lower part combining means 23 outputs data D3 (i) obtained by combining the upper part data U2 (i) and the lower part data L3 (i).

  The common redundant part adding means 24 outputs data D4 (i) obtained by adding the common redundant part to the data D3 (i) based on the run length information CRB2 of the common redundant part. The data D4 (i) with the common redundant portion added has the same bit length as the data D1 (j, i) input to the common redundant portion removing means 12 in FIG. Further, as described above, when the run length of the common redundant part is (WL-HB) or more, the upper part data U2 (i) is not output, so that the common redundant part is not output with respect to the lower part data L3 (i). Data D4 (i) with a common redundant part added based on the run length information CRB2 is output.

  The correlation component restoration unit 25 performs the inverse conversion process of the correlation component removal unit 11 of FIG. 1 on the data D4 (i) to which the common redundant portion is added based on the auxiliary information DI that specifies the correlation removal characteristic, and outputs the output sample. OUT (i) is restored and output.

  In addition, the function of each component of the lossless compression coding apparatus according to the first embodiment can be realized on a computer or a digital signal processor (DSP) by a software program.

  FIG. 5 is a software processing flowchart when the lossless compression encoding method of the present invention is realized on a DSP. The series of software processing shown in FIG. 5 is executed in units of blocks each including a predetermined number N of input samples IN (i) (0 <= i <N) divided at predetermined time intervals. First, initialization is performed at the head of the block (S1). Next, for the N samples of the block (S2), data from which the correlation component is removed is obtained according to the correlation removal characteristic. In FIG. 5, a primary difference value D1 (1, i) and a secondary difference value D1 (2, i) are calculated. At the same time, the input sample IN (i) is set in D1 (0, i) (S3). Next, for the calculated D (j, i) (0 <= j <= 2, 0 <= i <N), the run length information CRB (j) of the common redundant part is detected, and the common redundant part is detected. Are removed (S4).

  Next, for D (0, i), D (1, i), D (2, i) with the common redundant portion removed (S5), CRB (j) based on each CRB (j) Is smaller than (WL-HB) obtained by subtracting the minimum value (HB) of the code length HLN of the Huffman codeword of the Huffman coding table from the bit length (WL) of the input sample, the Huffman is separated after separating the upper part. Encoding processing is performed to obtain Huffman encoded data H1 (k, j, i) (S6, S7). Here, k is an index that specifies the Huffman coding algorithm.

  Next, the lower part is separated based on CRB (j) (S8). Next, the optimum combination of the correlation removal characteristic and the Huffman coding algorithm for outputting the data with the maximum compression efficiency is selected (S9). Finally, the auxiliary information DI specifying the selected correlation removal characteristic, the auxiliary information HI specifying the Huffman coding algorithm, the corresponding common redundant run length information CRB (DI), and the data are superimposed on the bit stream and output. (S10). Thereby, the same processing as the lossless compression coding apparatus shown in FIG. 1 is performed.

  Note that the software processing of the lossless compression decoding method according to the first embodiment of the present invention is performed in the reverse procedure of the lossless compression coding method of FIG. 5 and the output samples are restored from the bitstream. Omitted.

  As described above, according to the lossless compression coding method and apparatus of the present invention, when the input sample level is small, the data from which the common redundant part is removed is used as the lower part data, and the minimum of the Huffman codeword is set. By superimposing data smaller than the code length on the bitstream and not superimposing auxiliary information specifying the Huffman encoding algorithm on the bitstream, the compression efficiency can be improved, and compression is performed more than in the prior art. A lossless compression encoding method and apparatus with improved efficiency can be realized. This makes it possible to extend the recording time on a recording medium having a limited capacity.

Configuration diagram of lossless compression coding apparatus according to Embodiment 1 of the present invention Explanatory drawing showing the removal of a common redundant part in Embodiment 1 of this invention Explanatory drawing showing the effect of the information amount reduction in Embodiment 1 of this invention Configuration diagram of lossless compression decoding apparatus in Embodiment 1 of the present invention Software processing flow diagram of the lossless compression encoding method in Embodiment 1 of the present invention Configuration diagram of lossless compression coding apparatus of the prior art Correlation component removal unit internal configuration diagram Internal configuration diagram of common redundant part removal means Explanatory drawing showing removal of common redundant part Explanatory drawing showing separation of upper part and lower part Diagram showing Huffman coding table Explanatory drawing showing Huffman coding Explanatory drawing showing removal of common redundant part in the prior art Explanatory drawing showing prior art Huffman coding Configuration diagram of lossless compression decoding apparatus of the prior art Explanatory drawing showing Huffman decoding of prior art and upper part / lower part combination

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Lossless compression coding apparatus 2 Lossless compression decoding apparatus 11 Correlation component removal means 12 Common redundant part removal means 13 Upper part / lower part separation means 14 Huffman coding means 15 Maximum efficiency data output means 16 Multiplexer 21 Demultiplexer 22 Huffman decoding means 23 upper part / lower part combining means 24 common redundant part adding means 25 correlation component restoring means

Claims (14)

Correlation component for removing a correlation component between samples in units of a block consisting of a predetermined number of samples divided by a predetermined time from sample data expressed in a predetermined bit length and input in chronological order A removal step;
A common redundant part removing step for removing a common redundant part common to each sample of a bit length corresponding to the minimum value of the redundant run length of each sample in the block from the data after removing the correlation component;
When the bit length of the data from which the common redundant portion is removed is equal to or smaller than the minimum bit length of data obtained by Huffman coding, the Huffman coding is not performed, and the bit length of the data from which the common redundant portion is removed Is larger than the minimum bit length of data obtained by the Huffman coding, a Huffman coding step for obtaining Huffman coded data by performing Huffman coding on the data from which the common redundant portion has been removed, and
A lossless compression encoding method comprising: a superimposing step of superimposing the run length information of the common redundant part and the data from which the common redundant part is removed for each sample or the Huffman encoded data for each sample on a bitstream. Furthermore, it comprises an upper part / lower part separation step for separating the data from which the common redundant part is removed into upper part data of a predetermined bit length and remaining lower part data,
When the bit length of the data from which the common redundant part is removed is larger than the minimum bit length of data obtained by Huffman coding, the Huffman coding step obtains Huffman coded data by performing Huffman coding on the higher order data. The lossless compression encoding method according to claim 1, wherein in the superimposing step, the lower-order data is superimposed on the bitstream. The Huffman encoding step includes a plurality of Huffman encoding algorithms;
Further, when the bit length of the data after removing the common redundant portion is larger than the minimum bit length of the data obtained by Huffman coding of all the Huffman coding algorithms, the plurality of Huffman coding algorithms A maximum efficiency algorithm selection step for selecting a Huffman coding algorithm that maximizes the compression efficiency within the block,
The lossless compression encoding method according to claim 1, wherein in the superimposing step, a parameter indicating the selected Huffman encoding algorithm is further superimposed on the bitstream. In the superimposing step, the parameter is not superimposed on the bit stream when the bit length of the data from which the common redundant portion is removed is equal to or less than the minimum bit length of data obtained by Huffman coding of all the Huffman coding algorithms. The lossless compression encoding method according to claim 3. Correlation component for removing a correlation component between samples in units of a block consisting of a predetermined number of samples divided by a predetermined time from sample data expressed in a predetermined bit length and input in chronological order Removal means;
A common redundant part removing means for removing a common redundant part common to each sample having a bit length corresponding to the minimum value of the redundant run length of each sample in the block from the data after removing the correlation component;
When the bit length of the data from which the common redundant portion is removed is equal to or smaller than the minimum bit length of data obtained by Huffman coding, the Huffman coding is not performed, and the bit length of the data from which the common redundant portion is removed Is larger than the minimum bit length of the data obtained by the Huffman coding, Huffman coding means for obtaining Huffman coded data by Huffman coding the data from which the common redundant portion is removed,
A lossless compression encoding apparatus comprising: run length information of the common redundant portion; and a multiplexer that superimposes data obtained by removing the common redundant portion for each sample or Huffman encoded data for each sample on a bit stream. Furthermore, the data obtained by removing the common redundant part is provided with upper part / lower part separating means for separating the upper part data of a predetermined bit length and the remaining lower part data,
When the bit length of the data from which the common redundant part is removed is larger than the minimum bit length of data obtained by Huffman coding, the Huffman coding step obtains Huffman coded data by performing Huffman coding on the higher order data. The lossless compression encoding apparatus according to claim 5, wherein, in the superimposing step, the low-order data is superimposed on the bitstream. The Huffman encoding means has a plurality of Huffman encoding algorithms,
Further, when the bit length of the data after removing the common redundant portion is larger than the minimum bit length of the data obtained by Huffman coding of all the Huffman coding algorithms, the plurality of Huffman coding algorithms , Comprising a maximum efficiency algorithm selection means for selecting a Huffman coding algorithm that maximizes the compression efficiency within the block,
The lossless compression encoding apparatus according to claim 5, wherein the multiplexer further superimposes a parameter indicating the selected Huffman encoding algorithm on the bitstream. The multiplexer does not superimpose the parameter on the bitstream when the bit length of the data from which the common redundant portion is removed is equal to or less than the minimum bit length of data obtained by Huffman coding of all the Huffman coding algorithms. The lossless compression encoding apparatus according to claim 7. 2. The lossless compression encoding method according to claim 1, wherein a bitstream in which run length information of the common redundant portion and data obtained by removing the common redundant portion for each sample or Huffman encoded data for each sample is superimposed is input. Lossless compression decoding method,
An extraction step for extracting the run length information of the common redundant portion and the data obtained by removing the common redundant portion for each sample or the Huffman encoded data for each sample from the bitstream;
A Huffman decoding step of performing Huffman decoding on the extracted Huffman encoded data for each sample and outputting Huffman decoded data;
When the bit length obtained by subtracting the bit length of the extracted run length information of the common redundant portion from the predetermined bit length of the sample is larger than the minimum bit length of the data obtained by the Huffman coding, the Huffman decoding step A common redundant part is added to the Huffman-decoded Huffman decoded data based on the extracted run length information of the common redundant part, and the bit length of the extracted run length information of the common redundant part is subtracted from the predetermined bit length of the sample. When the extracted bit length is equal to or less than the minimum bit length of the data obtained by the Huffman coding, the extracted common redundant portion for each sample is removed, and the data is shared based on the run length information of the extracted common redundant portion. A common redundant part adding step for adding a redundant part;
A lossless compression decoding method comprising: a correlation component restoration step of restoring an original sample by adding a correlation component between the samples to the data to which the common redundant part is added. The bitstream is further superimposed with lower part data obtained by removing upper part data having a bit length corresponding to the codeword length of the Huffman coding from data obtained by removing the common redundant part for each sample,
In the extraction step, the lower part data is extracted,
Further, the Huffman decoded data is an upper part, and the extracted lower part data is used as a lower part, and an upper part / lower part combining step for combining them is provided.
In the common redundant part adding step, when the bit length obtained by subtracting the bit length of the extracted run length information of the common redundant part from the predetermined bit length of the sample is larger than the post-code length of the Huffman coding, Instead of the Huffman decoded data subjected to Huffman decoding in the Huffman decoding step, a common redundant part is added to the data combined in the upper part / lower part combining step based on the run length information of the extracted common redundant part The lossless compression decoding method according to claim 9. Huffman encoding algorithm information indicating one of a plurality of Huffman encoding algorithms is further superimposed on the bitstream,
In the extraction step, the Huffman coding algorithm information is extracted,
10. The lossless compression decoding method according to claim 9, wherein in the Huffman decoding step, Huffman decoding is performed with a Huffman encoding algorithm based on the extracted Huffman encoding algorithm information. 6. The lossless compression coding apparatus according to claim 5, wherein a bit stream in which run length information of the common redundant portion and data obtained by removing the common redundant portion for each sample or Huffman encoded data for each sample is superimposed is input. Lossless compression decoding device,
Extraction means for extracting the run length information of the common redundant portion and the data obtained by removing the common redundant portion for each sample or the Huffman encoded data for each sample from the bitstream;
Huffman decoding means for performing Huffman decoding on the extracted Huffman encoded data for each sample and outputting Huffman decoded data;
When the bit length obtained by subtracting the bit length of the extracted run length information of the common redundant portion from the predetermined bit length of the sample is larger than the minimum bit length of the data obtained by the Huffman coding, the Huffman decoding step A common redundant part is added to the Huffman decoded Huffman decoded data based on the extracted run length information of the common redundant part, and the bit length of the extracted common redundant part run length information is subtracted from the predetermined bit length of the sample. If the extracted bit length is equal to or less than the minimum bit length of the data obtained by the Huffman coding, the extracted common redundant portion for each sample is removed, and the data is shared based on the run length information of the extracted common redundant portion. Common redundant part adding means for adding a redundant part;
A lossless compression decoding apparatus comprising: a correlation component restoring unit that restores an original sample by adding a correlation component between the samples to the data to which the common redundant portion is added. The bitstream is further superimposed with lower part data obtained by removing upper part data having a bit length corresponding to the codeword length of the Huffman coding from data obtained by removing the common redundant part for each sample,
The extraction means extracts the lower part data,
Furthermore, the Huffman decoded data is an upper part, and the extracted lower part data is used as a lower part.
The common redundant part addition is performed when the bit length obtained by subtracting the extracted bit length of the run length information of the common redundant part from the predetermined bit length of the sample is larger than the post-code length of the Huffman coding. Instead of the Huffman decoded data that has been Huffman-decoded by the decoding means, a common redundant part is added to the data combined by the upper part / lower part combining means based on the run length information of the extracted common redundant part The lossless compression decoding apparatus according to claim 12. Huffman encoding algorithm information indicating one of a plurality of Huffman encoding algorithms is further superimposed on the bitstream,
The extraction means extracts the Huffman coding algorithm information,
13. The lossless compression decoding apparatus according to claim 12, wherein the Huffman decoding means performs Huffman decoding with a Huffman encoding algorithm based on the extracted Huffman encoding algorithm information.

Images