Video data compression.

Description: TECHNICAL FIELD [Detailed description of the invention]

【0001】

[Industrial application field]

The present invention relates to data coding.

【0002】

[Conventional technology]

Channel coding is used in digital magnetic recording devices such as digital video tape recorders (DVTRs) to match the frequency spectrum of the data to be recorded to the frequency response of the recording medium.

【0003】

For example, the previously proposed DVTR uses a channel code (so-called 8:16 code) to map an 8-bit data word to a 16-bit code word before the code word is recorded on a magnetic tape medium.

【0004】

The code word generated by the channel code is the maximum run length of a single binary value (for controlling the low frequency content of the encoded data), for controlling the high frequency content of the encoded data. ) Follow some rules regarding the minimum run length of a single binary value and the balance between 1 and 0 (to control the dc content of the encoded data).

【0005】

In the 8:16 code, each of the 256 possible 8-bit datawords has two complementary values corresponding to the 16-bit code for recording on tape media.

【0006】

This is possible 2¹⁶It means that there are 512 valid 16-bit codes out of (65536) permutations.

【0007】

[Problems to be Solved by the Invention]

In the previously proposed DVTR described above, if a 16-bit reproduction group does not correspond to one of 512 valid codes (for example, reproduction data error), the code is simply assumed to be error-prone. Is flagged.

【0008】

In addition, DPCM (Differential Code Modulation) is used to compress data, but in differential coding, if a data error occurs in the flow of encoded data, all the data values reconstructed after that are incorrect. There is a drawback that it ends up.

【0009】

In view of the above points, it is an object of the present invention to provide a coding method and an apparatus capable of efficiently detecting and correcting an error.

【0010】

[Means for solving problems]

In order to solve the above problems, the present invention provides a digital signal decoding method including the following means. That is, it corresponds to an effective data string that gives the lowest humming distance among the humming distances calculated between an input data string having an n-bit code length and a group of valid data strings having the same code length as the input data string. In a digital signal decoding device that outputs m-bit length data (n is larger than m) as decoded data for the input data string, if there are multiple valid data strings that give the minimum humming distance, is the minimum humming distance 1? When the minimum humming distance is determined to be 1 in the first means for determining whether or not, and in the first means above, the m-bit length data corresponding to any one of the plurality of valid data strings is decoded. If it is determined in the first means above that the minimum humming distance is not 1, the m-bit length data corresponding to the plurality of valid data strings having the smallest distribution length is used as the decoded data. A digital signal decoding apparatus is provided, which comprises the means of the above.

【0011】

In the above digital signal decoding apparatus, the second means is among the plurality of effective data strings having the minimum distribution length when there are a plurality of the plurality of valid data strings having the minimum distribution length. Provided is a digital signal decoding apparatus characterized in that m-bit length data corresponding to any one of the above is used as decoding data.

【0012】

The present invention also provides a digital signal decoding method including the following means. That is, it corresponds to an effective data string that gives the lowest humming distance among the humming distances calculated between an input data string having an n-bit code length and a group of valid data strings having the same code length as the input data string. In the digital signal decoding method that outputs m-bit length data (n is larger than m) as decoded data for the input data string, if there are multiple valid data strings that give the minimum humming distance, is the minimum humming distance 1? If the minimum humming distance is determined to be 1 in the first step of determining whether or not, and in the first step above, the m-bit length data corresponding to any one of the plurality of valid data strings is decoded. If it is determined in the first means above that the minimum humming distance is not 1, the m-bit length data corresponding to the plurality of valid data strings having the smallest distribution length is used as the decoded data. Provided is a digital signal decoding method comprising the steps of.

【0013】

Further, in the above digital signal decoding method, in the second step, when there are a plurality of valid data strings having the minimum distribution length, the plurality of valid data strings having the minimum distribution length are present. Also provided is a digital signal decoding method characterized in that m-bit length data corresponding to any one of the above is used as the decoding data.

【0014】

【0015】

【0016】

Assuming that the error results from a short burst error, this method selects the valid n-bit code word in which the bits that differ from the current n-bit group are closest to each other.

【0017】

【0018】

【0019】

【0020】

【0021】

【0022】

【0023】

【0024】

【0025】

【0026】

【0027】

【0028】

【0029】

【0030】

【0031】

【0032】

【0033】

【0034】

【0035】

【0036】

【0037】

【0038】

【0039】

【0040】

【0041】

[Example]

Referring to FIG. 1, a block diagram of a digital video tape recorder is shown, where the input video signal 10 is data compressed and block formatted before being recorded on the tape medium 20.

【0042】

Similarly, the data recovered from the tape medium 20 is complemented and stretched before being output as the output video signal 30.

【0043】

The input video signal 10 is first fed to the frequency uncorrelator 40, which contains a bank of horizontal and vertical finite impulse response filters, divides the data and evolves into 10 subband components in the 2D spatial frequency domain. Each frequency-converted image including.

【0044】

The data representing the frequency-converted image is supplied from the frequency uncorrelator 40 to the quantizer, where the data is subjected to lossy quantization.

【0045】

The quantized data is sent to the differential pulse code modulation (DPCM) encoder 60 in the form of a 12-bit data sample, the operation of which is further described below.

【0046】

The data from the DPCM encoder 60 is sent again to the entropy encoder 70 in the form of a 12-bit data sample. This encoder employs run-length and Huffman coding to perform data compression.

【0047】

The output of the entropy encoder 70 includes a data word having a length that varies in such a way that the data pattern that occurs more commonly in the data output by the DPCM encoder 60 is represented by a shorter data word.

【0048】

The variable length data word output by the entropy encoder 70 is temporarily stored in the record buffer 80 before being formatted into a fixed length "internal" data block by the block formatter and the ECC encoder 90.

【0049】

The block formatter and ECC encoder 90 also assign an error correction code ECC to each internal data block. This code can be used when playing back to detect data errors in its internal data block.

【0050】

The block formatter and ECC encoder supply the data start signal to the DPCM encoder 60 and the entropy encoder 70. The purpose of this data start signal will be described below.

【0051】

The formatted internal data blocks output by the block formatter and ECC encoder 90 are sent to the channel encoder 100 where the contiguous 8-bit datawords of the internal data blocks become the corresponding 16-bit codewords. It is encoded.

【0052】

For this reason, the coding performed by the channel encoder 100 is referred to as the 8:16 code. Since channel coding is performed, the coded data has more preferable properties for recording on a magnetic medium.

【0053】

For example, the magnetic medium does not respond to the recording of the DC dc component. This becomes a problem when digital data has long-lasting 1s or long-lasting 0s.

【0054】

Similarly, the reproduction of the high frequency components of the signal becomes unreliable due to the limited frequency response of the normal reproduction channel.

【0055】

Therefore, the 8-bit data word of the internal data block has a maximum transition width (maximum number of bits between adjacent transitions of encoded data) and a minimum transition width (minimum number of bits between adjacent transitions of encoded data). ) Is used to transfer to a 16-bit code that is compelled to follow.

【0056】

The 16-bit code words output by the channel coder 100 are recorded in series along the helical track of the magnetic tape medium 20.

【0057】

The data reproduced from the magnetic tape medium 20 is sent to the channel decoder 110 and the synchronous detector 120. The synchronous detector 120 detects the timing of the reproduced data and generates a data strobe signal that gives a clock pulse corresponding to each data word of the reproduced data.

【0058】

The data synchronization detector 120 also examines the reproduced data, detects synchronization information indicating the start of each of its internal data blocks, and a block strobe signal containing a series of strobes or clock impulses that mark the start of each internal data block. Is output.

【0059】

The generation of the block and data strobe signals by the synchronous detector 120 is performed using a "flywheel" mechanism, whereby the block and data strobe signals are constantly generated despite occasional replay data errors. ..

【0060】

This block and data strobe signal is supplied to the reproduction buffer and the ECC decoder 130, the entropy decoder 140, and the DPCM decoder 150.

【0061】

The operation of the channel decoder 110 is described in more detail below, but basically the channel decoder is a corresponding 8-bit that forms a continuous internal data block of 16-bit code words reproduced from the tape medium 20. It works to convert back to data.

【0062】

This internal data block is subject to ECC decoding by the playback buffer and ECC decoder 130, and an error flag is set in association with the input data block to indicate whether each internal data block has been reproduced quite successfully.

【0063】

The playback buffer and the ECC decoder 130 also perform error correction of playback errors for a short period of time. Internal data blocks that are not flagged as error, that is, blocks that have been successfully regenerated or blocks that have been successfully corrected by the replay buffer and ECC decoder, are subject to entropy decoding by the entropy decoder 140 and DPCM decoding by the DPCM decoder 150. Will be done.

【0064】

The data output by the DPCM decoder 150 is sent to the inverse quantizer 160, where the operation opposite to the quantization operation performed by the quantizer 50 is performed.

【0065】

This dequantized data is sent to the error concealer 165. This concealer conceals any data errors that were not corrected by the replay buffer and ECC decoder. Further, from there, it is sent to the frequency correlator 170, which performs complementary processing with the processing of the frequency uncorrelator 40.

【0066】

FIG. 2 shows the frequency-separated image data output by the frequency uncorrelator 40.

【0067】

In FIG. 2, the increase in horizontal frequency in the two-dimensional spatial frequency domain is represented along the horizontal axis, and the increase in vertical frequency is represented along the vertical axis.

【0068】

If the expression in Fig. 2 is applied to the right side and the lower side, it corresponds to increasing the spatial frequency.

【0069】

Figure 2 shows the input video signal converted into seven spatial frequency subbands. This is done in the frequency uncorrelator 40 by horizontal and vertical complementary low and high pass filtering to divide the data into four primary subbands of the same dimensions in the spatial frequency domain.

【0070】

Of these four equidimensional subbands, those representing the lowest horizontal and vertical spatial frequencies are subject to further horizontal and vertical filtering and division to form the four second subband components.

【0071】

The second subband, which represents the lowest horizontal and vertical spatial frequencies, will be cited as the dc (direct current) subband 200. It occupies 1/16 of the area of the spatial frequency domain.

【0072】

Since various subbands can have different degrees of quantization by the quantizer 50 based on this spatial frequency domain, in general, the data representing the high frequency components of this two-dimensional spatial frequency domain are subject to more stringent quantization. Will be done.

【0073】

FIG. 3 is a circuit diagram of the DPCM encoder 60. The encoder includes a subtractor 62, an adder 64 and a delay unit 66 connected to a feedback circuit.

【0074】

In this embodiment, since a 12-bit wide signal is supplied to the DPCM encoder, the adder 64 and the subtractor 62 are modulo 2.¹²(Modulo 4096) Operates according to arithmetic.

【0075】

(Generally, modulo 2 for b-bit width signals^{b b}Arithmetic is used to ensure that the DPCM code generated by the DPCM encoder 60 has the same number of bits as the signal supplied to that DPCM encoder).

【0076】

Each 12-bit sample received by the DPCM encoder 60 is first sent to subtractor 62, in which case the output of delay unit 66 is currently subtracted from the input 12-bit sample (modulo 4096).

【0077】

The output of the subtractor 62 is supplied to the entropy encoder 70 as a 12-bit output sample and is also added to the output of the delay unit 66 by the adder 64 (modulo 4096) to regenerate the input sample.

【0078】

This regenerated input sample is delayed by the delay unit 66 for one sample period and subtracted from the next input sample received by the subtractor 62 (modulo 4096).

【0079】

The clear (CLR) signal applied to this delay unit 66 can be used to set the output of that delay unit 66 to zero.

【0080】

If this output is set to zero, the value subtracted by the subtractor 62 from the current input sample will be zero, which will be sent directly to the entropy encoder 70 without the current input sample being subject to differential coding. Means that.

【0081】

DPCM coding is especially useful when applied to data with similar or identical continuous values. In this case, the increase in compression performed by the entropy encoder 70 can be achieved by applying DPCM coding to the relatively highly correlated data of the dc (direct current) subband 200.

【0082】

DPCM coding does not apply to the rest of the subband. Therefore, for the data from these remaining subbands, the CLR input of the delay unit 66 is permanently set, forcing the output of the delay unit 66 to zero.

【0083】

For example, during shuttle (fast) playback of the data stored on the tape medium 20, only some of the internal data blocks are successfully reproduced.

【0084】

The first data sample of each internal data block is not subject to DPCM coding, instead, so that each internal data block may be decoded independently of the adjacently recorded internal data blocks in this situation. , Non-DPCM and included in the internal data block.

【0085】

This sets the CLR input of delay unit 66 (thus the output of delay unit 66 to zero) for the data start signal from the block formatter and ECC encoder 90 and the first sample to be contained in each internal data block. Is achieved by applying to.

【0086】

This behavior is shown in the following table.

[table 1]

<img file="JPP3348741B2_D0001.tif" />【0087】

This table shows the 4-bit DPCM coding applied to one block of 12 data samples received by the DPCM encoder 60.

【0088】

When the first data sample (8) is received, the CLR input of the delay unit 66 is set and the first sample is sent directly to the entropy encoder 70 without applying differential coding.

【0089】

For subsequent samples, the CLR input is not set, so the value of the previous data sample is subtracted from this sample to generate a DPCM output sample.

【0090】

For example, the second sample received has a value of 4, which is then subtracted by the subtractor 62 with modulo 16 (in a 4-bit system). This subtraction is performed in modulo 16 arithmetic, so the corresponding DPCM sample is created as follows: DPCM = 4-8 = -4; -4 MOD 16 = 12 [0091]

In the embodiment shown in FIG. 1, data compression is applied separately to the luminance (Y) and chrominance (C) components of the input video signal 10 to generate an internal data block with a luminance data section and a chrominance data section. To do.

【0092】

The luminance data section and chrominance data section vary in length between contiguous internal data blocks, but the sum of their lengths is constant.

【0093】

The luminance and chrominance data sections represent the corresponding parts of the respective luminance and chrominance frequency-separated images generated by the frequency uncorrelator 40.

【0094】

FIG. 4 shows the DPCM decoder 150, which includes a modulo 4096 adder 152 and a delay unit 154. The output of this delay unit can be set to zero by applying a CLR signal to the delay unit 154.

【0095】

Each data sample received by the adder 152 is added (modulo 4096) to the output of the delay unit 154 to reconstruct the original sample as supplied to the DPCM encoder 60.

【0096】

The output of delay unit 154 represents a pre-reconstruction data sample. In this way, the delay unit 154 and the adder 152 operate in the opposite manner to the operations of the delay unit 66 and the subtractor 62 of the DPCM decoder 60.

【0097】

The CLR input to the delay unit 154 is for received data corresponding to all spatial frequency subbands except the dc (direct current) subband, and the first (non-DPCM) of the luminance or chrominance data section of each internal data block. Coded) Used to set the output of delay unit 154 to zero for a sample. This latter operation is performed by applying a block strobe generated by the sync detector 120 to the CLR input of the DPCM decoder 150.

【0098】

FIG. 5 shows the internal data blocks generated as described above. The total length of the internal data block 210 is fixed between consecutive internal data blocks. However, the relative dimensions of the luminance data section 220 and the chrominance data section 230 can be varied based on the information content of the luminance and chrominance components of the input video signal.

【0099】

The internal data block 210 includes a header 240, which contains a pointer indicating the boundary between the luminance data section 220 and the chrominance data section 230.

【0100】

Following the header is the Luminance Data Section 220, which contains the first PCM Luminance Sample 250 (ie, the non-differentiatedly encoded sample) followed by several Luminance DPCM Samples 260 as described above.

【0101】

Similarly, the chrominance data section 230 includes a first chrominance PCM sample 270 followed by several chrominance DPCM samples 280.

【0102】

An error correction code 290 is attached to the internal data block, and it is used for correcting minor reproduction errors of the internal data block and detecting all other errors.

【0103】

FIG. 6 is a block diagram of the channel coder 100. The channel encoder 100 is a 16-bit code supplied in series for recording each 8-bit word of each internal data block generated by the block formatr and ECC encoder 90 on a tape medium 20. Convert to word.

【0104】

The channel encoder 100 includes a channel encoder programmable read-only memory (PROM) 102, which is a register provided for feedback of the two most significant bits (MSB) of the 16-bit code generated by the PROM 102. It has.

【0105】

A 10-bit address is formed using the 8-bit data word supplied to the PROM 102 and the 2-bit fed back from the 16-bit code, and the next 16-bit code to be output to that address is stored.

【0106】

2-possible 8-bit input signal⁸Each of the street bit patterns is associated with different complementary pairs of 16-bit codewords according to the codebook stored in PROM102. However, only one codeword in each complementary pair needs to be stored in the PROM.

【0107】

Each of the 16-bit codewords is 2¹⁶Choose from a possible 16-bit combination of streets. Therefore, each codeword has the following conditions: Maximum transition width equal to or less than 1.6 bits; 2. Maximum run length equal to or less than 5 bits at the beginning and end; Minimum between first and last transitions equal to or greater than 3.2 bits Transition width; and 4.1 and 0 are equal numbers, i.e. dc balance.

【0108】

Thus, for each input 8-bit word, there are two possible 16-bit codes, one of which is the other binary complement. Which codeword in the pair is the output depends on the pre-output codeword as described below.

【0109】

These 16-bit codewords are recorded in series on magnetic tape medium 20 in a manner that records the least significant (LSB) bits first.

【0110】

The current 8 using the two most significant bits of each 16-bit codeword to maintain coercion to the minimum and maximum transition widths across the boundary between adjacently recorded 16-bit codewords. Determine which of the codeword complementary pairs corresponding to the bit data words should be used.

【0111】

For example, if the two MSBs of the previous codeword were 00 and the 16-bit codeword corresponding to this 8-bit input dataword was 10, then the previous and current dataword recordings were recorded on a magnetic tape medium. The pattern is 0010.

【0112】

This does not comply with the minimum transition width coercion matched above. Therefore, the 16-bit code complement corresponding to the current 8-bit input data word is used. As a result, bit pattern 0001 is recorded across the boundary between the two codewords.

【0113】

This 16-bit codeword is output in parallel by the PROM 102 and converted to serial for recording by the 16-bit serializer 104.

【0114】

FIG. 7 shows the Hamming distance between valid 16-bit codewords generated by the channel coder 100.

【0115】

The Hamming distance between two n-bit words is equal to the number of different bits in the corresponding positions in the two words. The Hamming distance may be evaluated for two specific n-bit words by counting the number of logical "1" s in one word formed by the exclusive OR of two n-bit words.

【0116】

For example, the Hamming distance for the following two words is<img file="JPP3348741B2_D0002.tif" />Is.

【0117】

The 16-bit code by the channel coder 100 is chosen so that the Hamming distance between the valid codes is 3 for at least the majority (90%) of the codes. This feature is used in this embodiment, in which a 16-bit word is reproduced from the tape and fed to the channel decoder 10, which is inconsistent with the valid 16-bit code but differs by a Hamming distance of 1 from the valid 16-bit code. Then, the code reproduced by mistake can be corrected to a valid code separated by a Hamming distance of 1 with a high probability. However, in the above explanation, it is assumed that only one bit is incorrect during reproduction. Such corrections are illustrated by arrow 300 in FIG.

【0118】

The high percent (about 90%) correction of single-bit errors described with reference to FIG. 7 is only part of the massive correction process performed by the channel decoder 100. 8 and 9 are flowcharts showing the operation of the channel decoder 110.

【0119】

In FIGS. 8 and 9, the 16-bit code reproduced from the tape medium 20 is received by the channel decoder 110 in step 310.

【0120】

In step 320, the received 16-bit code and the valid code 2 corresponding to all possible 8-bit data that could have been supplied to the channel encoder 110.⁸The Hamming distance between each pair is determined.

【0121】

Since half of the valid 16-bit sign (256) is the complement of the other half, we only need to consider the Hamming distance up to 8.

【0122】

This means that for any code that has a Hamming distance greater than 8 from the currently received code, the complement of that code (which is another valid code) has a Hamming distance of 16-h from the currently received code. Because.

【0123】

In step 330, one or more valid codes with the lowest Hamming distance from the currently received code are selected for post-processing.

【0124】

If only one of the valid codes is selected in step 330 (determined in step 340), in step 350 the 8-bit data word corresponding to the selected valid code is output by the channel decoder 110. Processing is completed for the received 16-bit code.

【0125】

This corresponds to an error-free receive code (in which case the single Hamming distance would be zero), or the situation illustrated in FIG. 7, where the following is assumed. That is, only single-bit errors occur, and the regenerated 16-bit code with many errors should be replaced with the nearest valid code.

【0126】

If, in step 340, more valid codes than 1 should be selected, in other words, it is determined that the currently received 16-bit code is equidistant in Hamming distance from 2 or more valid 16-bit codes. Then, in step 360, it is determined whether the Hamming distance from those two or more nearest codes is greater than one.

【0127】

If not, the channel decoder outputs an 8-bit dataword corresponding to any one of the two or more nearest codes in step 370 and ends the post-processing accordingly.

【0128】

However, if the Hamming distance between the currently received code and the two or more nearest valid codes selected in step 330 is greater than 1, then step 380 determines the distribution length for the selected code. To.

【0129】

The distribution length for each of the two or more selected valid signs is between the first and last logical value "1" in one word created by the exclusive OR of the selected valid signs for the currently received 16-bit sign. Is equal to the bit separation of.

【0130】

For example, the result of the sum of the exclusive ORs used in the calculation of the Hamming distance of the received code from two valid 16-bit codes is the following code. Hamming distance distribution length 0001011000000000 3 3 And 0010000000100100 3 11 [0131]

At step 390, a decision is made as to whether a single sign of two or more selected valid signs has the lowest distribution length.

【0132】

If so, in step 400, the 8-bit word corresponding to the selected valid code with the lowest distribution is output, and then the processing of the current 16-bit code is finished.

【0133】

If, in step 390, it is determined that two or more of the selected codes have equal low distribution lengths, then an 8-bit word corresponding to any one of the selected valid codes having the same minimum distribution length will At step 410, the output is completed and the processing of the current 16-bit code is completed.

【0134】

FIG. 10 is a block diagram showing an embodiment of the channel decoder 110. The reproduced data from the medium is loaded into the 16-bit latch circuit 112 under the control of the blocks and data strobes used to indicate the start of each 16-bit code word.

【0135】

The contents of the latch circuit 112 are supplied as an address input to the PROM 114. The PROM 114 is programmed according to a series of steps shown in FIGS. 8 and 9.

【0136】

Therefore, 2¹⁶For any of the possible received 16-bit code words, the proper 8-bit data word that would be generated by the steps in FIGS. 8 and 9 is an address corresponding to the 16 bits of that code word. It is accumulated in the place of. The output of PROM114 is a parallel 8-bit data stream containing continuous 8-bit data words.

【0137】

[Effect of the invention]

According to the data coding method of the present invention, code errors can be detected and corrected efficiently.

[Simple explanation of drawings]

[Figure 1]

It is a block diagram of a digital video tape recorder.

[Figure 2]

It is a diagram which shows the division of a frequency-separated image data.

[Fig. 3]

It is a block diagram of a DPCM encoder.

[Fig. 4]

It is a block diagram of a DPCM decoder.

[Fig. 5]

It is a diagram which shows the formatted internal block.

[Fig. 6]

It is a block diagram of a channel coder.

[Fig. 7]

It is a diagram which shows the Hamming distance between the effective code output by the channel coder of FIG.

[Fig. 8]

It is a flowchart which shows the operation of a channel decoder.

[Fig. 9]

It is a flowchart which shows the operation of a channel decoder.

[Fig. 10]

It is a block diagram of a channel decoder.

[Explanation of symbols]

20 Magnetic recording medium 110 channel decoder 120 synchroscope 130 Playback buffer and decoder 140 Entropy Decoder 150 DPCM decoder 160 Inverse quantizer 165 Error concealer 170 Frequency Correlator 30 Video data output