WO2018159105A1 - 画像復号化装置 - Google Patents

画像復号化装置 Download PDF

Info

Publication number
WO2018159105A1
WO2018159105A1 PCT/JP2018/000260 JP2018000260W WO2018159105A1 WO 2018159105 A1 WO2018159105 A1 WO 2018159105A1 JP 2018000260 W JP2018000260 W JP 2018000260W WO 2018159105 A1 WO2018159105 A1 WO 2018159105A1
Authority
WO
WIPO (PCT)
Prior art keywords
code data
unit
data
length code
variable length
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/JP2018/000260
Other languages
English (en)
French (fr)
Japanese (ja)
Inventor
晃一 古谷
真吾 柴田
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shikino High Tech Co Ltd
Original Assignee
Shikino High Tech Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shikino High Tech Co Ltd filed Critical Shikino High Tech Co Ltd
Priority to US16/489,237 priority Critical patent/US11251808B2/en
Publication of WO2018159105A1 publication Critical patent/WO2018159105A1/ja
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • H—ELECTRICITY
    • H03—ELECTRONIC CIRCUITRY
    • H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M7/00—Conversion of a code where information is represented by a given sequence or number of digits to a code where the same, similar or subset of information is represented by a different sequence or number of digits
    • H03M7/30—Compression; Expansion; Suppression of unnecessary data, e.g. redundancy reduction
    • H03M7/40—Conversion to or from variable length codes, e.g. Shannon-Fano code, Huffman code, Morse code
    • H—ELECTRICITY
    • H03—ELECTRONIC CIRCUITRY
    • H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M7/00—Conversion of a code where information is represented by a given sequence or number of digits to a code where the same, similar or subset of information is represented by a different sequence or number of digits
    • H03M7/30—Compression; Expansion; Suppression of unnecessary data, e.g. redundancy reduction
    • H03M7/46—Conversion to or from run-length codes, i.e. by representing the number of consecutive digits, or groups of digits, of the same kind by a code word and a digit indicative of that kind
    • H—ELECTRICITY
    • H03—ELECTRONIC CIRCUITRY
    • H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M7/00—Conversion of a code where information is represented by a given sequence or number of digits to a code where the same, similar or subset of information is represented by a different sequence or number of digits
    • H03M7/30—Compression; Expansion; Suppression of unnecessary data, e.g. redundancy reduction
    • H03M7/60—General implementation details not specific to a particular type of compression
    • H03M7/6005—Decoder aspects
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N1/00—Scanning, transmission or reproduction of documents or the like, e.g. facsimile transmission; Details thereof
    • H04N1/41—Bandwidth or redundancy reduction
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/42—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals characterised by implementation details or hardware specially adapted for video compression or decompression, e.g. dedicated software implementation
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/90—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using coding techniques not provided for in groups H04N19/10-H04N19/85, e.g. fractals
    • H04N19/91—Entropy coding, e.g. variable length coding [VLC] or arithmetic coding
    • H—ELECTRICITY
    • H03—ELECTRONIC CIRCUITRY
    • H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M7/00—Conversion of a code where information is represented by a given sequence or number of digits to a code where the same, similar or subset of information is represented by a different sequence or number of digits
    • H03M7/30—Compression; Expansion; Suppression of unnecessary data, e.g. redundancy reduction
    • H03M7/40—Conversion to or from variable length codes, e.g. Shannon-Fano code, Huffman code, Morse code
    • H03M7/42—Conversion to or from variable length codes, e.g. Shannon-Fano code, Huffman code, Morse code using table look-up for the coding or decoding process, e.g. using read-only memory
    • H03M7/425—Conversion to or from variable length codes, e.g. Shannon-Fano code, Huffman code, Morse code using table look-up for the coding or decoding process, e.g. using read-only memory for the decoding process only

Definitions

  • the present invention relates to an image decoding apparatus for decoding image data (code data) compressed using a Huffman code.
  • the image encoding device includes a DCT unit, a quantization unit, a zigzag scan unit, and a Huffman encoding unit.
  • the DCT unit divides image data converted from RGB color data into YUV luminance color difference data into 8 ⁇ 8 pixel blocks, and for each of the divided blocks, the image data corresponding to each pixel has a frequency component ( Conversion to DCT (referred to as “Discrete Cosine Transform” coefficient) is performed to generate coefficient data composed of one DC (Direct Current) coefficient and 63 AC (Alternating Current) coefficients.
  • the quantization unit converts each 8 ⁇ 8 coefficient data (numerical value) generated for each block by the DCT unit according to each numerical value corresponding to a previously prepared quantization table (8 ⁇ 8 numerical value). Each of them is divided and quantized. Further, the zigzag scanning unit linearly arranges the 8 ⁇ 8 coefficient data quantized for each block in the quantization unit by scanning zigzag from the DC coefficient to the AC coefficient ( That is, a process of generating 64 coefficient data (one-dimensionally arranged) is performed, and the Huffman encoding unit uses the Huffman code for each piece of coefficient data arranged one-dimensionally by the zigzag scanning unit. A process for generating and outputting encoded data encoded in this manner is performed.
  • FIG. 7 shows a data structure of general code data (JPEG stream) generated by this image encoding device.
  • the code data includes a front marker group composed of various markers, encoded compressed image data, and an EOI (End (Of Image) marker that defines the end of the code data. It has an arranged structure.
  • an SOI Start Of Image
  • DQT Define Quantization Table
  • SOS Start Of Scan
  • the compressed image data represents each coefficient data for each block by a Huffman code obtained by encoding zero-run information and group number information, and a variable-length code having an additional bit representing a coefficient value as a unit.
  • the variable-length code data is continuously connected.
  • the image decoding apparatus 100 includes a Huffman decoding unit 110, an inverse zigzag scanning unit 130, an inverse quantization unit 140, and an IDCT unit 150 (see Patent Document 1 below).
  • the Huffman decoding unit 110 generally includes a code input unit 111, an internal buffer 112, a front marker analysis unit 113, a final marker analysis unit 114, a code analysis unit 115, and a frequency as shown in FIG.
  • the conversion unit 125 is configured.
  • the code input unit 111 performs processing for receiving the code data input from the outside and storing the code data in the internal buffer 112. From the internal buffer 112, a front marker analysis unit 113, a final marker analysis unit 114, and a code analysis unit Code data is transmitted to each of 115.
  • the front marker analysis unit 113 extracts the front marker group included in the code data, transmits the extracted front marker group to the inverse zigzag scanning unit 130, and the final marker analysis unit 114 A process of detecting an EOI marker in the code data transmitted from the buffer 112 and transmitting the detection signal to the inverse zigzag scanning unit 130 is performed.
  • the code analysis unit 115 includes an input control unit 116, a code word detection unit 117, a zero run number / group number recognition unit 118, and an additional bit output unit 119.
  • the input control unit 116 controls the input of code data input from the internal buffer 112. Specifically, the input control unit 116 sequentially inputs the data related to the marker, and the maximum data that can be taken by the variable length code data that is one unit for the variable length code data after the SOS marker. A data amount slightly exceeding the amount is input, and a shift signal corresponding to the bit length of the variable-length code data processed by the codeword detection unit 117, the zero run number / group number recognition unit 118, and the additional bit output unit 119 is input. Transmit to the internal buffer 112.
  • the internal buffer 112 deletes data (processed variable length code data) corresponding to the received shift signal from the beginning of the stored data, and the next variable length code.
  • the data is cued by shifting it to the beginning.
  • the internal buffer 112 has a sufficient storage capacity for storing a plurality of variable-length code data, and deletes the processed variable-length code data to find the next variable-length code data. Then, subsequent code data is sequentially input from the outside to the internal buffer 112 under the control of the code input unit 111.
  • the codeword detection unit 117 recognizes (detects) a codeword, that is, data related to a Huffman code for each variable-length code data based on a Huffman code table recognized from a DHT marker in the input front marker group. Then, a process of transmitting the recognized Huffman code data and the variable length code data to the zero run number / group number recognition unit 118 is performed.
  • the zero run number / group number recognition unit 118 recognizes the zero run number and group number defined by the Huffman code based on the Huffman code data detected by the codeword detection unit 117, and recognizes the recognized zero run number and group number.
  • the variable length code data is output to the additional bit output unit 119.
  • the additional bit output unit 119 recognizes the additional bit length corresponding to the group number based on the data related to the group number received from the zero run number / group number recognition unit 118, and from the input variable length code data.
  • the data corresponding to the additional bit length after the Huffman code data is recognized as an additional bit, and the data related to the zero run number and the data related to the recognized additional bit are output to the frequency converter 125.
  • the additional bit output unit 119 outputs a signal related to the bit length of the processed variable length code data, that is, the bit length obtained by adding the additional bit length to the bit length of the Huffman code detected by the codeword detection unit 117.
  • the input control unit 116 transmits the shift signal corresponding to the bit length of the variable-length code data to the internal buffer 112 as described above.
  • the frequency conversion unit 125 generates quantized coefficient data (DCT coefficient data) in a one-dimensional array for each block based on the data relating to the number of zero runs transmitted from the code analysis unit 115 and the additional bits. The data is restored and transmitted to the reverse zigzag scanning unit 130.
  • DCT coefficient data quantized coefficient data
  • the marker information of the front marker group included in the code data is analyzed by the marker analysis unit 113 and transmitted to the inverse zigzag scanning unit 130, and then the code analysis unit 115 and the frequency conversion unit 125 restore the one-dimensional array coefficient data from the compressed image data for each block and transmit it to the inverse zigzag scanning unit 130, which processes all the code data and the EOI marker as the final marker analysis
  • the marker information of the EOI marker is transmitted from the final marker analysis unit 114 to the inverse zigzag scanning unit 130.
  • the inverse zigzag scanning unit 130 performs an inverse process on the coefficient data decoded by the Huffman decoding unit 110 to the above zigzag scanning, thereby converting the coefficient data of the one-dimensional array into an 8 ⁇ 8 two-dimensional array. Performs processing for conversion into array coefficient data.
  • the inverse quantization unit 140 outputs the 8 ⁇ 8 coefficient data output from the inverse zigzag scanning unit 130 and transmits the coefficient data defined by the DQT marker that is also transmitted from the inverse zigzag scanning unit 130.
  • the IDCT (Inverse Discrete Cosine Transform) unit 150 performs a process of converting into coefficient data before quantization by multiplying each corresponding numerical value in the quantization table.
  • the coefficient data converted by the inverse quantization unit 140 Is converted into 8 ⁇ 8 pixel YUV luminance color difference data, and the YUV luminance color difference data is finally converted into RGB color data.
  • the Huffman codes for defining the number of zero runs and the group number have 12 group numbers of 0 to B for DC coefficients and zero run numbers of 0 to F for AC coefficients.
  • the combination of 16 pieces and 10 pieces of group numbers 1 to A plus the combination of zero run number 0 and group number 0, and zero run number F and group number 0 is 162, for a total of 174 pieces Necessary.
  • the codeword detection unit 117 of the conventional image decoding apparatus 100 described above detects the Huffman code data from the variable length code data, there are 174 data groups positioned sequentially from the beginning of the variable length code data.
  • a match detection process for sequentially comparing the Huffman code data and detecting a data group that matches the Huffman code data is repeatedly executed 174 times.
  • the appearance frequency of Huffman codes obtained by encoding coefficient data in units of 8 ⁇ 8 pixel blocks is equal to all Huffman codes.
  • the frequency of appearance of a plurality of Huffman codes is high.
  • the matching detection process is performed for a Huffman code having a high appearance frequency
  • the Huffman code data in the variable-length code data can be detected with a very high probability with a short detection processing time.
  • the processing in the code analysis unit 115 of the Huffman decoding unit 110 needs to be realized within one clock, but the processing in the code analysis unit 115 Can be executed within one clock with a sufficient margin. Therefore, also in this sense, conventionally, the processing time in the code analysis unit 115 can be shortened.
  • the present invention has been made in view of the above circumstances, and provides an image decoding apparatus capable of reducing the decoding time as a whole by reducing the processing time in the Huffman decoding unit. For that purpose
  • the present invention for solving the above problems is as follows.
  • Compressed image data compressed by Huffman encoding processing having a structure in which Huffman code obtained by encoding zero-run information and group number information, and variable-length code data in units of additional bits following this are continuously connected
  • a Huffman decoding unit for decoding the compressed image data The Huffman decoding unit analyzes the variable length code data sequentially input, and sequentially recognizes and outputs zero run information and group number information related to the variable length code data, and the code analysis unit.
  • an image decoding device configured to include a frequency conversion unit that sequentially generates and outputs coefficient information
  • the code analysis unit includes: An input control unit for collectively inputting a plurality of variable length code data; The first variable length code data that has been input is processed without any restrictions, and the zero run information, group number information, and additional bit information are recognized, and the recognized zero run information and additional bit information are output to the frequency converter.
  • a primary analysis processing unit One or more variable length code data subsequent to the variable length code data processed by the primary analysis processing unit is processed, and when the variable length code data is variable length code data to be processed in advance, the variable length code data
  • the zero-run information, group number information and additional bit information related to the long code data are recognized, and the recognized zero-run information and additional bit information are output to the frequency conversion unit, while the variable-length code data is a variable to be processed. If it is not long code data, a process for discarding the data after the variable length code data is performed, and then a signal for inputting the data after the variable length code data for which the recognition process has been completed is sent to the input control unit.
  • the present invention relates to an image decoding apparatus including a continuous analysis processing unit for transmission.
  • variable-length code data is collectively input by the input control unit.
  • the first input variable-length code data is The zero run information, the group number information, and the additional bit information are recognized without being restricted, and the recognized zero run information and additional bit information are output to the frequency converter.
  • the primary analysis processing unit inputs the total of 174 Huffman codes including the 12 Huffman codes set for the DC coefficient data and the 162 Huffman codes set for the AC coefficient.
  • the Huffman code (code word) existing in the variable length code data is recognized by detecting which Huffman code matches the code data existing in the first variable length code data.
  • the primary analysis processing unit recognizes the number of zero runs and the group number (only the group number in the case of a DC coefficient) corresponding to the recognized Huffman code and adds the bit length from the recognized group number. Is recognized from this additional bit length, and each piece of information related to the recognized number of zero runs and additional bits is output to the frequency converter.
  • variable length code data subsequent to the variable length code data processed by the primary analysis processing unit is processed by the continuous analysis processing unit, and the variable length code data is processed in a predetermined variable length.
  • code data zero-run information, group number information and additional bit information related to the variable-length code data are sequentially recognized, and the recognized zero-run information and additional bit information are output to the frequency conversion unit. If the long code data is not the variable length code data to be processed, the data after the variable length code data is discarded.
  • the continuation analysis processing unit After executing these processes, the continuation analysis processing unit performs a process of transmitting a signal for inputting data after the variable length code data for which the recognition process has been completed to the input control unit.
  • variable-length code data to be processed in this continuous analysis processing unit is set empirically in advance.
  • the appearance frequency of Huffman codes obtained by coding coefficient data in units of 8 ⁇ 8 pixel blocks is not uniform in all Huffman codes, or The appearance frequency of a plurality of Huffman codes is high.
  • variable-length code data that is likely to be processed in the continuous analysis processing unit is variable-length code data related to the Huffman code having a high appearance frequency as described above. Therefore, the continuous analysis processing unit processes only the Huffman codes selected from those having a high appearance frequency, that is, variable length code data related to zero run information and group number information.
  • variable length code data if variable length code data is all processed, it is necessary to perform 174 matching detection processes in order to detect the Huffman code, but the processing target is related to a Huffman code having a specific appearance frequency.
  • variable-length code data By focusing on variable-length code data, the time required for the coincidence detection process for detecting the Huffman code can be shortened, and the time required for the process can be shortened without causing waste in the process. Can do.
  • variable-length code data that is not a processing target in the continuous analysis processing unit is processed again in the primary analysis processing unit again, so as to ensure that the compressed image data is reliably decoded,
  • the processing time can be shortened.
  • the input processing by the input control unit, the primary analysis processing unit, and the continuous analysis processing unit are repeatedly executed until the processing of a batch of compressed image data is completed.
  • coefficient information is sequentially generated based on the zero-run information and additional bit information sequentially output from the code analysis unit.
  • the continuous analysis processing unit is configured to sequentially process a plurality of variable length code data, and the variable length code data to be processed is limited as the subsequent variable length code data.
  • the variable-length code data to be processed may be configured to be gradually reduced in stages.
  • the image decoding apparatus is generally composed of an electronic circuit. In this case, a series of processes of an input control unit, a primary analysis processing unit, and a continuous analysis processing unit constituting the code analysis unit are performed within one clock. Although it is necessary to complete, if the time required for the coincidence detection processing in the continuous analysis processing unit is gradually shortened, the number of variable length code data that can be processed within one clock can be increased, and the processing time can be shortened. Can be planned.
  • the processing target in the continuous analysis processing unit, can be narrowed down to variable-length code data related to a specific high-frequency Huffman code.
  • the time can be shortened, and the time required for the processing can be shortened without causing waste of processing.
  • variable-length code data that is not a processing target in the continuous analysis processing unit is reliably processed again in the primary analysis processing unit, so that the compressed image data is reliably decoded while ensuring that The processing time can be shortened.
  • FIG. 1 is a block diagram showing a schematic configuration of an image decoding apparatus according to the present embodiment.
  • the image decoding apparatus 1 of the present example is different from the conventional image decoding apparatus 100 shown in FIG. 8 only in the configuration of the Huffman decoding unit 10, and an inverse zigzag scan.
  • the unit 130, the inverse quantization unit 140, and the IDCT unit 150 have the same configuration as that of the conventional image decoding device 100. Therefore, in FIG. 1, the same components are denoted by the same reference numerals, and detailed description thereof is omitted below.
  • the code data (JPEG stream) handled in this example has a data structure similar to that of the code data shown in FIG.
  • the Huffman decoding unit 10 of this example includes a code input unit 111, an internal buffer 112, a front marker analysis unit 113, a final marker analysis unit 114, a code analysis unit 20, and a frequency conversion unit 125.
  • the code input unit 111, the internal buffer 112, the front marker analysis unit 113, the final marker analysis unit 114, and the frequency conversion unit 125 are the same as those in the conventional image decoding device 100.
  • the same reference numerals are assigned and detailed description thereof is omitted.
  • the image decoding device 1 of this example is realized by hardware such as an electronic device provided with an appropriate electronic circuit for executing each process.
  • it can be realized by a general-purpose computer including a CPU, a ROM, a RAM, and the like.
  • the code input unit 111, the front marker analysis unit 113, the final marker analysis unit 114, the code analysis unit 20 and the frequency conversion unit 125, the zigzag scan unit 130, the inverse quantization unit 140, and the IDCT unit 150 Each function is realized by a computer program.
  • the code analysis unit 20 includes an input control unit 21, a primary analysis processing unit 31, a secondary analysis processing unit 41, a tertiary analysis processing unit 51, and a quaternary analysis processing unit 61.
  • the secondary analysis processing unit 41, the tertiary analysis processing unit 51, and the quaternary analysis processing unit 61 constitute a continuous analysis processing unit.
  • the input control unit 21 controls the input of code data input from the internal buffer 112. Specifically, the input control unit 21 sequentially inputs the data related to the marker, and the maximum data that can be taken by the variable length code data that is one unit for the variable length code data after the SOS marker. In consideration of the amount, the amount of data including at least four variable length code data is input.
  • the shift signal corresponding to the bit length of the variable-length code data processed by the primary analysis processing unit 31, the secondary analysis processing unit 41, the tertiary analysis processing unit 51, and the quaternary analysis processing unit 61 is sent to the internal buffer. 112.
  • the primary analysis processing unit 31 includes a codeword detection unit 32, a zero-run number / group number recognition unit 33, and an additional bit output unit 34.
  • the secondary analysis processing unit 41 includes a codeword detection unit 42, It consists of a zero run number / group number recognition unit 43 and an additional bit output unit 44.
  • the tertiary analysis processing unit 51 includes a code word detection unit 52, a zero run number / group number recognition unit 53, and an additional bit output unit 54.
  • the fourth analysis processing unit 61 includes a code word detection unit 62, It comprises a zero run number / group number recognition unit 63 and an additional bit output unit 64. Each part will be described below.
  • the codeword detection unit 32 based on the Huffman code table recognized from the DHT marker in the code data, for the input first variable length code data, data related to the Huffman code (codeword) (hereinafter, (“Huffman code data”) is detected.
  • the codeword detection unit 32 has 12 group numbers with respect to DC coefficients of 0 to B, 16 zero run numbers with respect to AC coefficients of 0 to F, and group numbers of 1 to A. 174 Huffman codes in total (combinations of 10 combinations of the above) plus a combination of zero run number 0, group number 0, and zero run number F, group number 0 are detected (recognized). ) Process is executed to recognize the Huffman code data present in the first variable length code data.
  • the codeword detection unit 32 transmits the detected Huffman code data and the plurality of input variable length code data to the zero run number / group number recognition unit 33.
  • the zero run number / group number recognition unit 33 recognizes the zero run number and the group number corresponding to the Huffman code data transmitted from the codeword detection unit 32, and then adds the bit length corresponding to the recognized group number. And the data related to the recognized zero-run number, the group number and the additional bit length, the data related to the bit length of the Huffman code transmitted from the codeword detection unit 32, and the plurality of variable length code data to the additional bit output unit The process which transmits to 34 is performed.
  • the additional bit output unit 34 converts the first variable length code data from the first variable length code data to the additional bit length based on the data related to the bit length of the Huffman code and the additional bit length transmitted from the zero run number / group number recognition unit 33. Data related to the corresponding additional bits is extracted.
  • the additional bit output unit 34 transmits the data related to the zero run number transmitted from the zero run number / group number recognition unit 33 and the data related to the extracted additional bits to the frequency conversion unit 125 and the first A process of transmitting data relating to the bit length of the variable-length code data and the second and subsequent variable-length code data from which the first variable-length code data is deleted to the codeword detection unit 42 is performed.
  • the codeword detection unit 42 determines whether the second variable length code data of the second and subsequent variable length code data transmitted from the additional bit output unit 34 is Huffman code data. Perform detection processing.
  • the codeword detection unit 42 performs grouping with a zero run number for the AC coefficient of 0, a group number of 0 to 7, a zero run number of 1, a group number of 1 to 3, and a zero run number of 2.
  • the matching detection processing is executed for 19 Huffman codes with a total of 19 and the group number 1, and the Huffman code data existing in the second variable length code data is recognized.
  • the codeword detection unit 42 converts the Huffman code data and the input second and subsequent variable length code data into the zero run number / It transmits to the group number recognition part 43.
  • the data related to the bit length of the first variable length code data transmitted from the additional bit output unit 34 is input to the input control unit. 21 and a process of discarding all input data. Thereby, the subsequent processing is stopped.
  • the input control unit 21 When the input control unit 21 receives data related to the bit length of the first variable length code data, the input control unit 21 transmits a signal related to the received bit length to the internal buffer 112 as a shift signal.
  • the next variable length code data is cued by deleting the data (bit variable length code data) corresponding to the received shift signal from the head of the stored data.
  • the zero run number / group number recognition unit 43 recognizes the zero run number and group number corresponding to the Huffman code data transmitted from the codeword detection unit 42, and then adds the bit length corresponding to the recognized group number. And the data related to the recognized zero run number, the group number and the additional bit length, the data related to the bit length of the Huffman code transmitted from the codeword detection 42, and the second and subsequent variable length code data are output as the additional bits. To the unit 44.
  • the additional bit output unit 44 converts the second variable length code data from the second variable length code data to the additional bit length based on the data relating to the bit length of the Huffman code and the additional bit length transmitted from the zero run number / group number recognition unit 43. Data related to the corresponding additional bits is extracted.
  • the additional bit output unit 44 transmits the data related to the zero run number transmitted from the zero run number / group number recognizing unit 43 and the data related to the extracted additional bits to the frequency conversion unit 125, and the first and Data relating to the total bit length of the second variable length code data and the third and subsequent variable length code data from which the second variable length code data has been deleted are transmitted to the codeword detection unit 52.
  • the codeword detection unit 52 determines the Huffman code data of the third variable length code data among the third and subsequent variable length code data transmitted from the additional bit output unit 44. Perform detection processing.
  • the codeword detection unit 52 performs grouping with a zero run number for the AC coefficient of 0, a group number of 0-7, a zero run number of 1, a group number of 1-2, and a zero run number of 2.
  • the coincidence detection process is executed for 13 types of Huffman codes in which the number is 1, the number of zero runs is 3, the group number is 1, the number of zero runs is 4, and the group number is 1, and there is present in the third variable length code data. Recognizes Huffman code data.
  • the codeword detection unit 52 converts the Huffman code data and the input third and subsequent variable length code data into the zero run number / It transmits to the group number recognition part 53.
  • the codeword detection unit 52 converts the Huffman code data and the input third and subsequent variable length code data into the zero run number / It transmits to the group number recognition part 53.
  • data relating to the total bit length of the first and second variable length code data transmitted from the additional bit output unit 44 is obtained. While transmitting to the said input control part 21, the process which discards all the input data is performed. Thereby, the subsequent processing is stopped.
  • the input control unit 21 When the input control unit 21 receives data related to the total bit length of the first and second variable length code data, the input control unit 21 transmits a signal related to the received bit length to the internal buffer 112 as a shift signal. In the buffer 112, the next variable length code data is cued by deleting the data of the bit length corresponding to the received shift signal from the head of the stored data.
  • the zero run number / group number recognition unit 53 recognizes the zero run number and group number corresponding to the zero run number and group number from the Huffman code data transmitted from the codeword detection unit 52, and then adds the number corresponding to the recognized group number. Recognizes the bit length, adds the data related to the recognized number of zero runs, the group number and the additional bit length, the data related to the bit length of the Huffman code transmitted from the codeword detection 52, and the third and subsequent variable length code data. It transmits to the bit output part 54.
  • the additional bit output unit 54 adds the third variable length code data from the third variable length code data based on the data related to the bit length of the Huffman code and the data related to the additional bit length transmitted from the zero run number / group number recognition unit 53. Data related to additional bits corresponding to the bit length is extracted.
  • the additional bit output unit 54 transmits the data related to the zero run number transmitted from the zero run number / group number recognizing unit 53 and the data related to the extracted additional bits to the frequency converting unit 125 and the first to Data relating to the total bit length of each third variable-length code data and the fourth and subsequent variable-length code data from which the third variable-length code data has been deleted are transmitted to the codeword detection unit 62.
  • the codeword detection unit 62 determines the Huffman code data of the fourth variable length code data among the fourth and subsequent variable length code data transmitted from the additional bit output unit 64. Perform detection processing.
  • the codeword detection unit 62 performs grouping with a zero run number for the AC coefficient of 0, a group number of 0 to 7, a zero run number of 1, a group number of 1 to 2, a zero run number of 2
  • the coincidence detection process is executed for 11 types of Huffman codes in total of 1, and Huffman code data existing in the fourth variable length code data is recognized.
  • the codeword detection unit 62 converts the Huffman code data and the input fourth and subsequent variable length code data into the zero run number / It transmits to the group number recognition part 63.
  • the codeword detection unit 62 converts the Huffman code data and the input fourth and subsequent variable length code data into the zero run number / It transmits to the group number recognition part 63.
  • data relating to the total bit length of the first to third variable length code data transmitted from the additional bit output unit 54 is obtained. While transmitting to the said input control part 21, the process which discards all the input data is performed. Thereby, the subsequent processing is stopped.
  • the input control unit 21 When receiving the data related to the total bit length of the first to third variable length code data, the input control unit 21 transmits the received bit length signal to the internal buffer 112 as a shift signal, and In the buffer 112, the next variable length code data is cued by deleting the data of the bit length corresponding to the received shift signal from the head of the stored data.
  • the zero run number / group number recognizing unit 63 recognizes the zero run number and the group number corresponding thereto from the Huffman code data transmitted from the code word detecting unit 62, and then adds the number corresponding to the recognized group number. Recognizes the bit length, adds the data related to the recognized zero-run number, the group number and the additional bit length, the data related to the bit length of the Huffman code transmitted from the codeword detection 62, and the fourth and subsequent variable length code data. Transmit to the bit output unit 64.
  • the additional bit output unit 64 adds the additional variable length code data from the fourth variable length code data based on the data related to the bit length of the Huffman code and the data related to the additional bit length transmitted from the zero run number / group number recognition unit 43. Data related to additional bits corresponding to the bit length is extracted. Then, the additional bit output unit 64 transmits the data related to the zero run number transmitted from the zero run number / group number recognition unit 63 and the data related to the extracted additional bits to the frequency conversion unit 125 to transmit the first to fourth data Data relating to the total bit length of the second variable length code data is transmitted to the input control unit 21 and a process of discarding all input data is performed.
  • FIG. 5 shows the number of zero runs and group numbers related to AC coefficients that appear when image data is compressed at a general compression rate of 20%, in other words, the appearance frequency of Huffman codes.
  • the appearance frequency of the number of zero runs and the group number is not uniform in all combinations, and the appearance frequency of the number of zero runs and the group number relating to a plurality of combinations is high.
  • the appearance frequency is high for the group numbers 0 to 7, and when the number of zero runs is 1, the group number is 1
  • the appearance frequency is high for 3 and the number of zero runs is 2 and 3
  • the appearance frequency is high for the group numbers 1 and 2, respectively, and when the number of zero runs is 4, 5, 6, and 7, the group numbers are respectively Appearance frequency is high for 1.
  • variable-length code data that is likely to be processed in the secondary analysis processing unit 41, the tertiary analysis processing unit 51, and the quaternary analysis processing unit 61 is related to the Huffman code having a high appearance frequency as described above.
  • Variable length code data Therefore, the secondary analysis processing unit 41, the third analysis processing unit 51, and the fourth analysis processing unit 61 select Huffman codes selected from those having a high appearance frequency, that is, variable lengths related to zero-run information and group number information. Only code data was processed.
  • variable length code data if variable length code data is all processed, it is necessary to perform 174 matching detection processes in order to detect the Huffman code, but the processing target is related to a Huffman code having a specific appearance frequency.
  • the processing target is related to a Huffman code having a specific appearance frequency.
  • the secondary analysis processing unit 41 sets the variable length code data related to 19 types of Huffman code data as a processing target, and the tertiary analysis processing unit 51 sets the variable length code data related to 13 types of Huffman code data as a processing target.
  • variable-length code data related to 11 types of Huffman code data is a processing target, and the processing targets are gradually narrowed down step by step, so the secondary analysis processing unit 41 to the fourth-order analysis processing unit. 61, the processing time is gradually shortened. Since the image decoding apparatus 1 of this example is generally composed of an electronic circuit, in this case, the processing from the input control unit 21 to the additional bit output unit 64 of the code analysis unit 20 is completed within one clock. There is a need to. Therefore, if the processing time is gradually shortened, the number of variable length code data that can be processed within one clock can be increased, and the processing time can be shortened.
  • variable length code data that is not processed in the secondary analysis processing unit 41 to the fourth analysis processing unit 61 is reliably processed again in the primary analysis processing unit 31, so that the compressed image data is processed.
  • the processing time can be shortened while ensuring that decoding is performed reliably.
  • code data is input from the internal buffer 112 under the control of the input control unit 21, and variable-length code data after the SOS marker is input.
  • an amount of data including at least four variable length code data is input at a time, and the input variable length code data is first transmitted to the primary analysis processing unit 31 for processing.
  • the code word detection unit 32 executes detection processing for 174 Huffman codes with respect to the first variable length code data, and the first variable length code data is included in the first variable length code data.
  • Existing Huffman code data is recognized.
  • the zero run number / group number corresponding to the recognized Huffman code data and the additional bit length corresponding to the group number are recognized by the zero run number / group number recognition unit 33.
  • the additional bit output unit 34 based on the data relating to the additional bit length, data relating to the additional bit corresponding to the additional bit length is extracted from the first variable length code data, and the data relating to the zero run number , And the data related to the extracted additional bits are transmitted from the additional bit output unit 34 to the frequency conversion unit 125, and the second and subsequent variable-length code data is the code word detection unit 42 of the secondary analysis processing unit 41. Sent to.
  • the code word detection unit 42 executes the detection process for the 19 Huffman codes described above for the second variable length code data, and when any of the 19 Huffman code data is detected.
  • the zero run number / group number recognition unit 43 recognizes the zero run number and group number corresponding to the next recognized Huffman code data, and the additional bit length corresponding to the group number.
  • the additional bit output unit 44 based on the data relating to the additional bit length, data relating to the additional bit corresponding to the additional bit length is extracted from the second variable length code data, and the data relating to the zero run number , And the data related to the extracted additional bits are transmitted to the frequency conversion unit 125, and the third and subsequent variable length code data are transmitted to the code word detection unit 52 of the tertiary analysis processing unit 51.
  • variable length code data is cued.
  • the code word detection unit 52 executes the detection process for the 13 Huffman codes described above for the third variable length code data, and when any of the 13 Huffman code data is detected.
  • the zero run number / group number recognition unit 53 recognizes the zero run number and group number corresponding to the next recognized Huffman code data and the additional bit length corresponding to the group number.
  • the additional bit output unit 54 based on the data relating to the additional bit length, data relating to the additional bit corresponding to the additional bit length is extracted from the third variable length code data, and the data relating to the zero run number And the data related to the extracted additional bits are transmitted to the frequency conversion unit 125, and the fourth and subsequent variable-length code data are transmitted to the codeword detection unit 62 of the quaternary analysis processing unit 61.
  • the total bit length of the first and second variable length code data transmitted from the additional bit output unit 44 is set. Such data is transmitted to the input control unit 21 and the subsequent processing is stopped. Then, the input control unit 21 transmits the received signal related to the bit length to the internal buffer 112 as a shift signal, and the internal buffer 112 deletes data corresponding to the received shift signal for the next time.
  • the variable length code data is cued.
  • the detection processing is executed for the 11 Huffman codes described above for the fourth variable length code data, and any one of the 11 Huffman code data is detected.
  • the zero run number / group number recognition unit 63 recognizes the zero run number and group number corresponding to the next recognized Huffman code data, and the additional bit length corresponding to the group number.
  • the additional bit output unit 64 based on the data relating to the additional bit length, data relating to the additional bit corresponding to the additional bit length is extracted from the fourth variable length code data, and the data relating to the zero run number The data relating to the extracted additional bits is transmitted to the frequency converter 125, and the data relating to the total bit length of the first to fourth variable length code data is transmitted to the input controller 21.
  • the total bit length of the first to third variable length code data transmitted from the additional bit output unit 64 is set. Such data is transmitted to the input control unit 21 and the subsequent processing is stopped.
  • the input control unit 21 transmits the received signal related to the bit length to the internal buffer 112 as a shift signal, and the internal buffer 112 deletes data corresponding to the received shift signal for the next time.
  • the variable length code data is cued.
  • the input control unit 21 is processed in the same manner until the processing of the batch of compressed image data is finished. And the primary analysis processing unit 31 to the fourth analysis processing unit 61 are repeatedly executed.
  • the secondary analysis processing unit 41, the tertiary analysis processing unit 51, and the quaternary analysis processing unit 61 are selected from those having a high appearance frequency. Since only Huffman codes, that is, variable length code data related to zero-run information and group number information are processed, detection of Huffman codes in the secondary analysis processing unit 41, the tertiary analysis processing unit 51, and the quaternary analysis processing unit 61 is performed. Processing can be performed in a short time.
  • FIG. 6A shows the processing in the code analysis unit 115 of the conventional image decoding apparatus 100
  • FIG. 6B shows the processing in the code analysis unit 20 of the image decoding apparatus 1 of this example.
  • the code analysis unit 115 of the conventional image decoding apparatus 100 needs to perform 174 matching detection processes in the detection process of the Huffman code data. It took a long time. Therefore, as shown in FIG. 6A, the processing (A) in the codeword detection unit 117, the processing (B) in the zero run number / group number recognition unit 118, the processing (C) in the additional bit output unit 119, and the input Even if the series of processes (D) in the control unit 116 can be processed within one clock, if a considerable remaining time occurs, the series of processes must be executed again within the remaining time. In this case, this remaining time becomes a complete dead time.
  • the primary analysis is performed to perform 174 Huffman code data coincidence detection processes within one clock as in the conventional case.
  • secondary analysis processing to quaternary analysis processing with a processing time much shorter than this, so that it is possible to process four variable length code data within one clock, and
  • the processing in the Huffman decoding unit 10 can be significantly shortened compared to the conventional case.
  • the time of the process A 2 in the code word detection unit 42 of the secondary analysis processing unit 41 is equal to the process A in the code word detection unit 32 of the primary analysis processing unit 31. 1 (equal to the conventional process A) is about 11/100, and the time of the process A 3 in the codeword detection unit 52 of the tertiary analysis processing unit 51 is about 7/100 of the time of the process A 1. Yes, the time of the processing A 4 in the code word detection unit 62 of the quaternary analysis processing unit 61 is about 6/10 of the time of the processing A 1 .
  • variable-length code data to be processed is gradually narrowed down from the secondary analysis processing unit 41 to the quaternary analysis processing unit 61. Since the processing time can be gradually shortened from the secondary analysis processing unit 41 to the fourth analysis processing unit 61, the number of variable length code data that can be processed within one clock can be increased more effectively. Further, the processing time can be shortened.
  • the relationship between the processing times is A 1 > A 2 > A 3 > A 4 .
  • variable length code data that is not processed in the secondary analysis processing unit 41 to the fourth analysis processing unit 61 is reliably processed again in the primary analysis processing unit 31, so that the compressed image data is processed.
  • the processing time can be shortened while ensuring that decoding is performed reliably.
  • the continuation analysis processing unit is composed of the three processing units of the secondary analysis processing unit 41, the tertiary analysis processing unit 51, and the quaternary analysis processing unit 61.
  • the number of units is not limited to this, and it is sufficient that the number of units is composed of one or more analysis processing units. In short, it is sufficient if the target time can be shortened.
  • the narrowing down of processing targets in the continuous analysis processing unit is not limited to the above example, It can be set as appropriate so as to shorten the time.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Compression, Expansion, Code Conversion, And Decoders (AREA)
  • Compression Or Coding Systems Of Tv Signals (AREA)
  • Compression Of Band Width Or Redundancy In Fax (AREA)
PCT/JP2018/000260 2017-02-28 2018-01-10 画像復号化装置 Ceased WO2018159105A1 (ja)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US16/489,237 US11251808B2 (en) 2017-02-28 2018-01-10 Image decoding device

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
JP2017035769A JP6821184B2 (ja) 2017-02-28 2017-02-28 画像復号化装置
JP2017-035769 2017-02-28

Publications (1)

Publication Number Publication Date
WO2018159105A1 true WO2018159105A1 (ja) 2018-09-07

Family

ID=63371186

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2018/000260 Ceased WO2018159105A1 (ja) 2017-02-28 2018-01-10 画像復号化装置

Country Status (3)

Country Link
US (1) US11251808B2 (cg-RX-API-DMAC7.html)
JP (1) JP6821184B2 (cg-RX-API-DMAC7.html)
WO (1) WO2018159105A1 (cg-RX-API-DMAC7.html)

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2002237754A (ja) * 2001-02-07 2002-08-23 Canon Inc 可変長復号化装置及び方法
JP2004179752A (ja) * 2002-11-25 2004-06-24 Kawasaki Microelectronics Kk ハフマンデコーダ
JP2011114525A (ja) * 2009-11-26 2011-06-09 Dainippon Printing Co Ltd 数値データ列の符号化/復号化の方法および装置

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP3292221B2 (ja) * 1993-09-14 2002-06-17 ソニー株式会社 画像圧縮符号化方法
JP3260925B2 (ja) 1993-09-28 2002-02-25 キヤノン株式会社 画像処理装置
WO1999044368A1 (fr) * 1998-02-27 1999-09-02 Kanebo Limited Dispositif de traitement de donnees d'images et procede de traitement
US6603413B2 (en) 2001-02-07 2003-08-05 Canon Kabushiki Kaisha Variable-length decoding apparatus and method
JP4434155B2 (ja) * 2006-02-08 2010-03-17 ソニー株式会社 符号化方法、符号化プログラムおよび符号化装置

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2002237754A (ja) * 2001-02-07 2002-08-23 Canon Inc 可変長復号化装置及び方法
JP2004179752A (ja) * 2002-11-25 2004-06-24 Kawasaki Microelectronics Kk ハフマンデコーダ
JP2011114525A (ja) * 2009-11-26 2011-06-09 Dainippon Printing Co Ltd 数値データ列の符号化/復号化の方法および装置

Also Published As

Publication number Publication date
JP6821184B2 (ja) 2021-01-27
JP2018142843A (ja) 2018-09-13
US11251808B2 (en) 2022-02-15
US20200014400A1 (en) 2020-01-09

Similar Documents

Publication Publication Date Title
JP6923677B2 (ja) ビデオ圧縮におけるエントロピーコーディングのための選択的ミキシング
RU2417518C2 (ru) Эффективное кодирование и декодирование блоков преобразования
US8588540B2 (en) Arithmetic encoding apparatus executing normalization and control method
US20120057799A1 (en) Run length coding with context model for image compression using sparse dictionaries
JP4522199B2 (ja) 画像符号化装置及び画像処理装置及びそれらの制御方法、並びにコンピュータプログラム及びコンピュータ可読記憶媒体
JP6070568B2 (ja) 特徴量符号化装置、特徴量符号化方法、及びプログラム
CN106341692A (zh) 图像数据的霍夫曼解码装置和方法
EP1946246A2 (en) Extended amplitude coding for clustered transform coefficients
US11265544B2 (en) Apparatus and method for image compression based on optimal sequential encoding scheme
KR102393743B1 (ko) 모드 심볼들을 사용하는 인코더, 디코더 및 방법
US20060039615A1 (en) Joint amplitude and position coding for photographic image and video coding
US8446299B2 (en) Method and device for encoding and decoding of data in unique number values
KR102219133B1 (ko) 컴팩트한 비트 평면(bit-plane) 데이터 압축을 위한 인코더, 디코더 및 부호화 방법
JP6821184B2 (ja) 画像復号化装置
JP2025145993A (ja) 復号装置及び復号方法
JP2019022040A (ja) 画像復号化装置
JP2018201143A (ja) 画像符号化装置
JP6821183B2 (ja) 画像復号化装置
JP6191173B2 (ja) 画像処理装置、方法及びプログラム
US20140205199A1 (en) Storage Area and Temporal Based Optimized Run Length Codec For Classic JPEG Image Compression (SAT-BORL CODEC)
CN107770558B (zh) Jpeg图像解码的方法、系统、装置及可读存储介质
JP2002016501A (ja) 可変長復号化装置
JPH0884260A (ja) 2次元画像データの圧縮方式および伸長方式
KR20150072853A (ko) 가변길이코딩을 이용한 인코딩 방법, 디코딩 방법, 및 그 시스템
JPH10163880A (ja) データ復号化器

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 18760304

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 18760304

Country of ref document: EP

Kind code of ref document: A1