KR20030052803A - UVLC Multiple Decoding Method - Google Patents

Abstract

PURPOSE: A method for multiple-extracting universal variable length codeword and a method for constructing a look-up table therefor are provided to speed up an extracting process by extracting a plurality of codewords at the same time. CONSTITUTION: The size of unit data to be processed in a bit stream at a time is set. An index is granted to the current unit data selected among the bit stream. All codewords are extracted from the current unit data, and the length from a first bit to a final bit in the rest of the bits is defined as a suffix-length. A binary value of the bit corresponding to the suffix-length is defined as a suffix-value. The length from a most significant bit to an extension bit of the codeword belonging to the previous unit data is defined as a prefix-length. A binary value of the bit corresponding to the prefix-length is defined as a prefix-value. The number of extractable codewords is defined as a number-of-codewords. The length and a binary value of each codeword are defined as codeword-lengths and codeword-values. The defined suffix-length, suffix-value, prefix-length, prefix-value, number-of-codewords, codeword-lengths and codeword-values are recorded at corresponding fields(21,22,23,24,25,26) of the index.

Description

Universal variable length code codeword multiple extraction method and look-up table construction method for it {UVLC Multiple Decoding Method}

The present invention relates to a Universal Variable Length Code (UVLC) codeword multiple extraction method and a look-up table construction method therefor, and more particularly, to an H.26L decoder. When decoding the UVLC codeword included in the bit stream input from (Decoder), UVLC codeword multi-extraction method and look-for for extracting a plurality of codewords at high speed by using the characteristics of UVLC It relates to how to configure up table.

H.26L follows the H.263 and MPEG-4 design of the next generation of international, currently driven by ITU-T SG16 Q.6, with the goal of designing high performance in a wide range of applications such as mobile applications and the Internet. Refers to the video coding standard.

The H.26L is currently in large part to show its performance as the next-generation international standard and is expected to be approved in 2002. H.26L includes advanced technologies compared to H.26x and MPEG-x technologies, which are the existing international video coding standards, and because of their complementary performance, H.26L provides superior performance compared to the existing coding standards. What is seen is reported.

In all international video code standard decoders, the process of distinguishing the codewords defined in the Variable Length Code (VLC) table from the received or retrieved bit stream is inevitably included. It is.

In H.26L, as in the existing coding standards, only one VLC table is used for encoding. The peculiarity is that the structure of the VLC table is regular, and this VLC table is UVLC in H.26L standardization. It is called.

The UVLC table 10 in H.26L consists of the respective codewords as shown in FIG. 1, as shown in the figure, where all codewords start with the first bit. It has an odd number of lengths and forms a structure in which the ends of all codewords are separated by the case where the odd bit of each codeword is '1'. Therefore, one codeword can be extracted by finding the point where the odd-numbered bit becomes '1' starting from the beginning of the bit string. If one codeword is found and the next bit is searched starting with the next bit, then the next codeword can be extracted, and this process is repeated to extract all codewords contained in the entire bit string. It is possible. In the figure, 'x' may have a value of '0' or '1' as an information bit.

However, such a conventional encoding extraction method requires bit shift processing, which can bring considerable computational load to most CPUs, and one code at a time. The only drawback is that it cannot be extracted.

The present invention has been made to solve the above problems, the object of which is to extract a plurality of codewords at the same time while eliminating the bit shift operation that burdens the central processing unit, UVLC codeword multiplexing It is an object of the present invention to provide an extraction method and a look-up table construction method therefor.

1 is a block diagram of a universal variable length code table,

2 is a diagram illustrating an embodiment of a look-up table configuration according to the present invention;

3 is a flowchart of a method for extracting a universal variable length code codeword according to an embodiment of the present invention.

※ Explanation of code for main part of drawing

10: UVLC table 20: look-up table

21: Suffix-length field 22: Suffix-value field

23: prefix-length field 24: prefix-value field

25: (internal) codeword-number 26: field for length and value of internal codeword

In order to achieve the above object, a method of constructing a look-up table for multiply extracting a universal variable length code (UVLC) codeword according to the present invention comprises: encoding codewords defined in a universal variable length code (UVLC) table in a bit string; A method of extracting, comprising: a first step of setting a unit data size to be processed at one time in the bit stream; A second step of assigning an index to the current unit data selected from the bit strings; A third step of defining a length from a first bit to a last bit as first information (Suffix-Length) after extracting all the codewords in the process of extracting a codeword from the current unit data; A fourth step of defining a binary value of a bit corresponding to the first information as second information (Suffix-Value); A fifth step of defining a length from a most significant bit of the current unit data to an extension bit of a codeword belonging to previous unit data as third information (Prefix-Length); A sixth step of defining a binary value of a bit corresponding to the third information as fourth information (Prefix-Value); A seventh step of defining the number of extractable codewords in the current unit data as fifth information (Number-of-Codeword); An eighth step of defining lengths (Codeword-Length) and binary values (Codeword-Value) for each codeword of the number defined in the fifth information as the sixth information; And a ninth step of recording the defined first to sixth information in a corresponding field of the index.

On the basis of the third information and the corresponding first information defined in the codeword extraction process from the previous unit data, the bit length of the most significant codeword in the codeword extraction process from the current unit data is defined, The bit length of the codeword is determined by adding the value of the first information to the value of the third information.

Further, based on the third information, the fourth information, and the corresponding second information defined in the codeword extraction process in the previous unit data, the most significant codeword in the codeword extraction process in the current unit data may be used. Define a binary value and write it to the corresponding field of the index,

The binary value of the most significant codeword may be obtained by, for example, adding the fourth information value to a value shifted bit by the number of bits of the third information value from the second information value, or as another example, the second value. to the product of the binary value of the information values and the value of ^K 2, obtained by adding the fourth information value, in which K represents the third information value.

As described above, the look-up table (LUT) according to the present invention distinguishes the case where the most significant bit of the current unit data is an even bit and an odd bit of the most significant codeword, respectively. It consists of an even lookup table and an odd lookup table.

In order to achieve the above object, a method for extracting a universal variable length code (UVLC) codeword according to the present invention is a decoding method for extracting codewords defined in a universal variable length code (UVLC) table from a bit string. Defines the length from the first bit to the last bit among the remaining bits as the first information (Suffix-Length) after extracting all the extractable codewords in the process of extracting codewords from the current unit data of the selected size of the corresponding bit string. The binary value of the bit corresponding to the first information is defined as the second information (Suffix-Value), and the length from the most significant bit of the current unit data to the extension bit of the codeword belonging to the previous unit data is determined by the third value. Define as information (Prefix-Length), define the binary value of the bit corresponding to the third information as the fourth information (Prefix-Value), and add in the current unit data The number of codewords that can be output is defined as fifth information (Number-of-Codeword), and the length (Codeword-Length) and binary value (Codeword-Value) for each codeword of the number defined in the fifth information is defined. A first step of defining sixth information; Reading the current unit data of the preset size from the input bit string, assigning an index of a current use table, and initializing the first information and the second information of a previous process of the previous unit data to '0'; Two steps; A third step of generating the defined third information value and obtaining the length of the first codeword by adding the generated third information value and the corresponding first information value of the previous unit data; The value of the first codeword is generated by generating the defined fourth information value and adding the fourth information value to a value shifted bit by the number of bits of the third information value from the corresponding second information value of the previous unit data. Obtaining a fourth step; And a fifth step of extracting a length and a value of a codeword following the first codeword based on the defined fifth information and sixth information.

In the codeword extraction process of the present invention as described above, the third to fifth steps may be performed in parallel, and the first information value and the second information value of the current unit data It is characterized by storing for codeword extraction.

According to the method of multi-extracting the UVLC codeword using the look-up table of the present invention as described above, it is possible to eliminate the bit shift operation and the bit comparison operation that are repeatedly performed in the existing UVLC codeword extraction process, The codeword can be extracted with

That is, the method proposed in the present invention processes all the bit strings in units of data (e.g., 1 byte or 1 word) that is optimal for processing by each system (e.g., decoder). Process them in batches.

The look-up table of the present invention uses an even look-up table using a structure in which odd bits of all codewords defined in the UVLC table are control bits and even bits are information bits. And odd look-up tables. In addition, each table includes a suffix-length as the first information, a suffix-value as the second information, a prefix-length as the third information, and the first information. Prefix-Value as 4 information and Number-of-Codewords as the fifth information are defined, and the length of the codeword by the number defined by the codeword number (Codeword-Length) ) And the value (Codeword-Value) are defined.

Using a look-up table having such a structure, according to the codeword extraction method of the present invention, at the beginning of the bit string, starting from an odd look-up table, the current status is continued with the unit data size. The bit stream is decoded into a predefined unit data unit by tracking and using an even look-up table or an odd look-up table according to the state of the tracked current unit data.

Hereinafter, a method of extracting a universal variable length code (UVLC) codeword according to a preferred embodiment of the present invention and a look-up table construction method thereof will be described in detail with reference to the accompanying drawings.

2 illustrates one embodiment of a look-up table configuration in accordance with the present invention.

The method proposed in the present invention can be applied by adjusting the size of the look-up table based on the data unit that can be optimally processed in the system (ie, decoder) to which the method is applied. Assuming that the unit of data to be processed at one time (hereinafter referred to as unit data) is N bits, the system generates a look-up table 20 having 2 ^N indexes (INDEX), as shown in FIG. Define two (that is, an even look-up table and an odd look-up table). Each index (INDEX 1 to INDEX 2 ^N ) has at least five variable number of fields defined, and the minimum five fields are Suffix-Length (21) and suffix-value ( Suffix-Value (22), Prefix-Length (23), Prefix-Value (24), and Number-of-Codewords (25). . And a field 26 for recording length and value of M codewords (Mwordwords: Codeword 1 to Codeword M), the M codewords having a start and end in unit data. Defined as an internal codeword that contains everything.

The even-look-up table and the odd-look-up table may be the case where the most significant bit (MSB) is the even-numbered bit of the codeword and the odd-numbered bit of the N-bit data as the current unit data to be respectively decoded. It is generated according to the result of the case.

In addition, the prefix and suffix designated by the index indicated by the current unit data to be processed are respectively the suffix field of the previous unit data in the previous extraction process and the subsequent extraction process. By adding to a value defined in the prefix field of the next unit data, it is defined to extract a codeword that spans two or more consecutive unit data.

That is, the values defined in each field of the look-up table of the present invention will be described in detail as follows.

Suffix-Length is the highest level after extracting all codewords extractable from the current unit data in the process of extracting the codeword from the current unit data (i.e., N bits of data). It is defined as the length from the L (1≤L≤N) th bit to the Nth bit which is the end of the unit data, which means that at least one bit or more is the corresponding codeword of the next processed data unit A bit length belonging to the first codeword) is defined as N-L + 1 '. If the Nth bit is the end of the extractable codeword, the Suffix-Length is zero.

Suffix-Value is L (1≤L≤N) which is the highest level after extracting all codewords extractable from the current unit data in the codeword extraction process in the current unit data. At least one bit from the nth bit to the Nth bit, which is the end of the unit data, belongs to the codeword of the next processing unit data, the Lth bit is the most significant bit and is the end of the current unit data. It is defined as the value of N-L + 1 bit string up to Nth bit. If the Nth bit is the end of the extractable codeword, the Suffix-Value is zero.

Prefix-Length means that in the process of extracting the codeword from the current unit data, if the K th bit starting from the most significant bit of the current unit data is an extension of the codeword belonging to the previous unit data, It is defined as K (1≤K≤N). In particular, this K value is added to the corresponding suffix-length generated in the codeword extraction process in the previous data unit, so that the first codeword in the codeword extraction process in the current data unit (ie, the most significant codeword or first). 1 codeword).

Prefix-Value means that, in the process of extracting the codeword from the current unit data, if the K-th bit starting from the most significant bit of the current unit data is an extension of the codeword belonging to the previous unit data, It is defined as a binary value of K (1≤K≤N) bits. Especially. This prefix-value is added to the suffix value generated in the process of extracting the codeword from the previous unit data and multiplied by 2 ^K , so that the first codeword extracted from the process of extracting the codeword from the current unit data (i.e., , The most significant codeword or the first codeword). On the other hand, another example of obtaining the value of the most significant codeword, the prefix- to a value shifted bit by the prefix-length from the suffix-value (binary value) generated during the codeword extraction process from the previous unit data Can be found by adding the values together.

Number-of-Codeword is defined as the number of codewords in which both the beginning and the end of the codeword are included in the current unit data composed of N bits as a single processing unit, so that the codeword can be extracted clearly. Although the codeword-number has the same index, its value is different in the case of an even look-up table and an odd look-up table.

Codeword-Length stores the length of the number of codewords specified in the codeword-number field, and Codeword-Value specifies the codeword-value in the codeword-number field. As many codeword values are stored.

The method of the present invention extracting the codeword from the UVLC table using the look-up table defined as above is as follows.

3 is a flowchart of a method for extracting a UVLC codeword according to an embodiment of the present invention. Since the look-up table of FIG. 1 is used, the table will be described with reference to the table.

First, in the initial process, the suffix-length and the suffix-value of the previous unit data as the previous process are initialized to '0' (S301), and the current use table is initialized to an odd look-up table (S302).

Then, N bits of unit data are read from the bit stream and used as an index of the currently used table. For example, when N is set to 6 bits in a bit string such as '001000 011100 001 ....', '001000' is read up to 6 bits based on the most significant bit of the bit string, and an odd look-up is used. Used as index 1 of the table (S303).

According to the definition of the look-up table described above with reference to FIG. 2, in the current unit data (hereinafter referred to as first unit data) '001000', the prefix-length is '0', the prefix-value is '0', The suffix-length is '3', the suffix-value is '0' and the internal codeword-number is 1, and the internal codeword according to the codeword-number is '001', the code for that codeword '001'. If the length is 3 and the signword-value is represented in binary, it is 1 in decimal.

Subsequently, the generated prefix-length and the suffix-length of the previous unit data are added to obtain a length of the first codeword (ie, a codeword that spans the current unit data and the previous unit data) extracted first. In step S201, the suffix-length of the previous unit data is initialized to '0', and the prefix-length is also '0' according to the definition of the look-up table, so that the length of the first codeword is '0'. . That is, the first codeword based on the prefix-length and the prefix value does not exist, and the codeword defined and extracted in the look-up table becomes the first codeword. This is the case when there is no first codeword (S304).

Subsequently, the prefix-value '0' generated according to the definition of the look-up table is converted into the suffix-value '0' and 2K (K is '0' as the prefix-length) of the previous unit data. In addition to the multiplied value, a value of the first codeword is obtained, and the result is '0' (S305).

That is, as a result of performing steps S204 and S205, since the length and the value of the first codeword that span the first unit data and the previous unit data are both '0', it is found that there is no corresponding codeword.

Next, '1', which is the value of the codeword-number, is read from the index 1 of the look-up table (S306), and the length of the inner codeword '3' and the internal codeword are performed by the codeword-number once. The value '001' is read through the index 1 (S307).

After the suffix-length '3' and the suffix-value '0' of the first unit data are stored in a corresponding field of the index 1, the codeword extraction process for the first unit data is finished. (S308) By repeatedly performing the steps S301 to S308, the codeword extraction process is continuously performed on the unit data, thereby extracting all the codewords defined in the UVLC from the input bit stream.

In the above description, only the codeword extraction process for the first unit data '001000', which is the initial 6-bit unit data in the bit string, has been described, and the current unit data '011100' as the 6-bit unit data thereafter (hereinafter, referred to as second unit data) is described. Referring to the codeword extraction process in the following).

First, in step S301, the suffix-length '3' and the suffix-value '0' of the first unit data stored in the corresponding field of the index 1 are read, and in step S302, the current usage table is determined. Since the most significant bit of the second unit data is the even bit of the codeword, an even look-up table is used here.

Subsequently, in step S303, 6-bit unit data '011100' as a bit string subsequent to the first unit data '001000' is read as second unit data and used as an index of an even look-up table that is a currently used table.

According to the definition of the look-up table described above with reference to FIG. 2, in the second unit data '011100', the prefix-length is '2', the prefix-value is '01' in binary, '1' in decimal, and the suffix. The length is '2', the suffix value is '0' and the inner codeword-number is '2', and the two inner codewords according to the codeword-number are the third bit in the second unit data '011100'. Is '1' and the fourth bit is '1'.

Then, in step S304, the first codeword extracted first by adding the generated prefix-length and the suffix-length of the previous unit data (that is, the second unit data that is the current unit data and the first unit data that is the previous unit data). The length of the codeword spanning between the suffixes), the suffix-length of the first unit data as the previous unit data is '3' and the prefix-length is '2' according to the definition of the look-up table. Since the length of the first codeword of the second unit data is '5'.

Then, the look-a value "1", the suffix of the first unit of data as the previous unit data - said prefix generated according to the definition of up table values 0 and 2 ^K (K is a prefix - the length The value of the first codeword is obtained by adding '2' to a binary operation), and the result is '00001' in binary and '1' in decimal.

That is, as a result of performing steps S304 and S305, the first codeword that spans between the first unit data and the first unit data in the second unit data that is the current unit data is the fourth unit of the first unit data from the fourth bit of the second unit data. It is extracted as '00001' as a code word spanning the second bit.

Next, in step S306, the value of the codeword-number of the second unit data is read through the corresponding index of the corresponding look-up table, and in step S307, the second unit data 'is twice the codeword-number. In 011100 ', the third bit' 1 'and the fourth bit' 1 'are extracted as codes, respectively, and the corresponding codeword length and codeword value are read through the specified index of the look-up table.

In operation S308, the suffix-length '2' and the suffix-value '2' of the second unit data are stored in a corresponding field of the index 1 to perform a codeword extraction process for the second unit data. After completion, the steps S301 to S308 are repeatedly performed to continuously perform the codeword extraction process from the unit data, thereby extracting all the codewords defined in the UVLC from the input bit stream.

In the entire process (steps S301 to S308) of the codeword extraction method of the present invention described above, the steps S304 to S308 can be performed simultaneously by using a parallel structure, so that multiple codewords can be extracted at the same time. Fast multiple codeword extraction is possible.

As described in detail above, a method for extracting a universal variable length code (UVLC) codeword according to the present invention and a method of constructing a look-up table therefor include a plurality of codewords while excluding bit shift operations that burden the central processing unit. Can be extracted at the same time, thereby speeding up the extraction process. In addition, the present invention proposes a fast codeword extraction technique of UVLC, which is a unique characteristic of H.26L as a next-generation international video coding standard, and can be used in all terminals of a next-generation communication system. In addition, since the look-up table can be adaptively designed and applied to the system according to the unit of data that can be optimally processed in the application system of the present invention, its use range is very widely applied.

Claims (10)

A method for extracting codewords defined in a universal variable length code (UVLC) table from a bit string, A first step of setting a unit data size to be processed at one time in the bit string; A second step of assigning an index to the current unit data selected from the bit strings; A third step of defining, as first information, a length from a first bit to a last bit among remaining bits after extracting all extractable codewords in the codeword extraction process in the current unit data; A fourth step of defining a binary value of a bit corresponding to the first information as second information; A fifth step of defining, as third information, a length from a most significant bit to an extension bit of a codeword belonging to previous unit data among the current unit data; A sixth step of defining a binary value of a bit corresponding to the third information as fourth information (Prefix-Value); A seventh step of defining the number of extractable codewords as fifth information in the current unit data; An eighth step of defining length and binary values for each codeword of the number defined in the fifth information as sixth information; And And a ninth step of recording the defined first to sixth information in a corresponding field of the index. The method of claim 1, On the basis of the third information and the corresponding first information defined in the codeword extraction process from the previous unit data, the bit length of the most significant codeword in the codeword extraction process from the current unit data is defined, And writing to the corresponding field of the index. 12. The method of claim 12, further comprising: extracting a universal variable length code (UVLC) codeword. The method of claim 2, The bit length of the most significant codeword is obtained by adding the value of the first information to the value of the third information, and configuring a look-up table for multiplexing of variable length code (UVLC) codewords. Way. The method according to claim 1 or 2, Based on the third information, the fourth information and the corresponding second information defined in the codeword extraction process in the previous unit data, the binary value of the most significant codeword in the codeword extraction process in the current unit data. And defining and writing the data in a corresponding field of the index. 11. The method of claim 12, further comprising: extracting the Universal Variable Length Code (UVLC) codeword. The method of claim 4, wherein The binary value of the most significant codeword is obtained by adding the fourth information value to a value shifted bit by the number of bits of the third information value from the second information value. UVLC) A Look-up Table Construction Method for Codeword Multiple Extraction. The method of claim 4, wherein To the binary value of the best single codeword is multiplied by the binary value of the associated second information value and the 2 ^K value value, obtained by adding the fourth values, where K is characterized in that the third information value A method of constructing a look-up table for extracting universal variable length code (UVLC) codewords. The method of claim 4, wherein The lookup-table includes an even lookup-table and an odd lookup-table, respectively, when the most significant bit (MSB) of the current unit data is an even bit and an odd bit of the most significant codeword. 12. A method for constructing a look-up table for universal variable length code (UVLC) codeword multiplexing. A decoding method for extracting codewords defined in a universal variable length code (UVLC) table from a bit string, After extracting all the extractable codewords in the process of extracting the codeword from the current unit data of the selected preset size of the bit string, the length from the first bit to the last bit among the remaining bits is defined as first information, and the first The binary value of the bit corresponding to the information is defined as the second information, and the length from the most significant bit of the current unit data to the extension bit of the codeword belonging to the previous unit data is defined as the third information. A binary value of a corresponding bit is defined as fourth information, the number of codewords that can be extracted in the current unit data is defined as fifth information, the length of each codeword of the number defined in the fifth information, and A first step of defining a binary value as sixth information; Reading the current unit data of the preset size from the input bit string, assigning an index of a current use table, and initializing the first information and the second information of a previous process of the previous unit data to '0'; Two steps; A third step of generating the defined third information value and obtaining the length of the first codeword by adding the generated third information value and the corresponding first information value of the previous unit data; The value of the first codeword is generated by generating the defined fourth information value and adding the fourth information value to a value shifted bit by the number of bits of the third information value from the corresponding second information value of the previous unit data. Obtaining a fourth step; And And a fifth step of extracting a length and a value of a codeword subsequent to the first codeword based on the defined fifth information and sixth information. Er multiple extraction method. The method of claim 8, And the third to fifth steps are performed in parallel with each other. The method of claim 8, And storing the first information value and the second information value of the current unit data for extracting codewords from the next unit data.