KR20090081277A - The apparatus for decoding the total zeros information with low power based on new reconstruction table and the method of thereof - Google Patents

Abstract

A decoding method and apparatus of low power total zero information based on a new decoding table are provided to offer an efficient memory structure storing total zero information, thereby reducing repeated memory access. Non-zero index quantity information showing the number of coefficients which are not zero is obtained. In case relation is simple among code words and total zero information corresponding to the non-zero index quantity information, the total zero information is calcualted(S402,S403). In case relation is not simple among the code words and the total zero information corresponding to the non-zero index quantity information, the total zero information is obtained by using a new decoding table made by utilizing the properties of the code words(S405~S410).

Description

The apparatus for decoding the total zeros information with low power based on new reconstruction table and the method of IEC}

The present invention relates to a method and apparatus for decoding total zero information based on a decoding table. More particularly, the present invention relates to restoring total zero information through an arithmetic operation. The present invention relates to a method and apparatus for restoring total zero information using a new decryption table created using a feature.

 The H.264 / AVC video compression standard, released by the ITU-T / ISO / IEC Joint Video Team in 2005, is a new standard that replaces MPEG-2 and MPEG-4. have. In Korea, where digital multimedia broadcasting (DMB), which is difficult to distinguish between broadcasting and communication, is emerging, the H.264 / AVC standard is adopted as a basic video technology for DMB service. Also, in the US and Europe, H.264 / AVC standard is being adopted as a video codec for next generation HD service, and competition for HD H.264 / AVC system technology is getting fierce. However, various new functions introduced in the H.264 / AVC standard have resulted in a significant increase in computation not only in encoding but also in decoding. For example, it is known that the calculation amount of the H.264 / AVC decoder is about four times that of the MPEG-2 decoder. The excessive computation amount of such a decoder is recognized as a major problem in developing competitive H.264 / AVC based multimedia services and products.

In the H.264 / AVC standard, transform coefficients are encoded and decoded by Context-based Adaptive Variable Length Coding (CAVLC) and Context-based Adaptive Binary Arithmetic Coding (CABAC). Currently, CAVLC is used in H.264 / AVC Baseline Profile and X Profile, and CAVLC and CABAC are used in Main Profile. However, in the H.264 / AVC Baseline Profile which mainly uses mobile video terminal services such as PMP, PDA, DMB and DVB-H, about 30% of the total decoding calculation amount is consumed by the CAVLC reconstruction unit. Therefore, for the development of a competitive H.264 / AVC Baseline Profile decoder, efficient implementation of the CAVLC reconstruction unit is of paramount importance.

The CAVLC decoder restores transform coefficients in a 4x4 block by the following four steps. First, T1s (Trailing_ones) and the number of non-zero coefficients (Tc) are used to determine the total number of transform coefficients (Coeff_token) with non-zero coefficients. The coefficients are obtained, and the third method determines the number of transform coefficients having a coefficient value of zero through total zero restore. Finally, by executing Run_before restoration, the actual position of transform coefficients in the 4x4 block is determined. These restoration steps require a lot of memory access because they are based on a variable length code table (hereinafter referred to as "VLCT").

To date, techniques have been developed for designing CAVLC reconstructors into VLSI using several implementation techniques that can reduce hardware cost and power consumption. However, since the CAVLC reconstruction unit is based on the reconstruction method proposed in the H.264 / AVC standard, there are many parts to be improved. Recently, a new decoding technique has been developed that uses the finite state machine (FSM) technique to restore run-before elements without using VLCT. In addition, an efficient Coeff_token reconstruction method has been proposed, which greatly reduces memory access by the newly designed VLCT.

However, since the restoration of the Total_Zeros syntax element is still based on the VLCT proposed in the H.264 / AVC standard, many memory accesses are made. 1 is a VLCT for a total zero information syntax element used in conventional H.264 / AVC. Due to the irregularity and complexity of the codeword, the VLCT of FIG. 1 is typically implemented in memory. Existing total zero information recursively accesses memory until a codeword matching the input bit string is found using the non-zero coefficient information value obtained in Coeff_Token decoding and a given input bit string. When the codeword matching the input bit string is found, the total zero information is recovered from the memory address. However, this repetition process is performed in units of bits, which causes a lot of memory access. In general, memory access is one of the major factors that increases the power consumption of the system and limits the overall operating speed of the system. In particular, the above problems are very important considerations when implementing a mobile video terminal or a multimedia service. Therefore, it is necessary to develop a low power total zero information decoding technique and apparatus through memory access reduction.

In order to solve the above problem, the present invention restores the total zero information through arithmetic operations, and when the arithmetic operations are impossible, total zeros using the new decoding table generated through structural feature analysis of the codeword existing in the existing VLCT. Its purpose is to provide a method and apparatus for restoring information.

In order to achieve the above object, the present invention provides a method for restoring total zero information based on a decoding table, the method comprising: obtaining non-zero coefficient number information (Tc) representing a total number of non-zero coefficients; Calculating the total zero information through an arithmetic operation when the relationship between the code words corresponding to the non-zero coefficient number information and the total zero information is simple; And when the relationship between the code words corresponding to the non-zero coefficient number information and the total zero information is not simple, acquiring the total zero information by using a decoding table generated by using the characteristics of the code words. It provides a total zero information recovery method based on a new decoding table comprising a.

The new decoding table does not include the existing codeword, but includes the total zero information.

If the relationship between the codeword and the total zero information is simple, the non-zero coefficient number information is 1, 14 or 15.

The calculating of the total zero information may be performed after obtaining zero bit number information Num_Z before the first bit 1 of codewords corresponding to the non-zero coefficient number information when the non-zero coefficient number information is 1; Reading the remaining bit information (iNFO) of a bit and reading two bits of input bit string information (IN [1: 0]) when the non-zero coefficient number information is 14 or 15; And calculating the total zero information through the arithmetic operation using information necessary according to the non-zero coefficient number information among the zero bit number information, the remaining bit information, and the input bit string information. The remaining bit information may indicate a bit present after the first bit 1 of the codewords.

The calculating of the total zero information may include obtaining bit string length information bit_rd_len using at least one of the input bit string information and the zero bit number information, and then using the bit string length information, And correcting the pointer.

Acquiring total zero information may include acquiring control information (i, j, k, m) using the non-zero coefficient number information; Obtaining a maximum possible value (max_Z_len) of the zero bit number information using the control information; Obtaining the zero bit number information from the bit string by the maximum possible value; Obtaining address information of the decoding table using at least one of the zero bit number information and the remaining bit information indicating the remaining bits present after the first bit 1 of the codewords; And selecting total zero information from the decryption table by using the address information.

Selecting total zero information may include reading upper 4-bit total zero information (U4) and lower 4-bit total zero information (L4) from the decoding table using the address information; And when the upper 4 bits total zero information and the lower 4 bits total zero information are not the same, after additionally reading 1 bit from an input bit string buffer, the upper 4 bits total zero information U4 and lower 4 bits total zero information ( Selecting one of L4) and selecting the upper 4 bit total zero information or the lower 4 bit total zero information when the upper 4 bit total zero information and the lower 4 bit total zero information are the same. do.

The acquiring of the total zero information may further include obtaining bit string length information bit_rd_len using the zero bit number information, and then correcting a pointer of an input bit string buffer using the bit string length information. It is characterized by.

In addition, in order to achieve the above object, the present invention provides a total zero information recovery apparatus based on a decoding table, generating control information (i, j, k, m) by using the non-zero coefficient number information A control information and a maximum possible value generator for generating a maximum possible value of the zero bit number information using the control information; A zero bit number information obtaining unit for obtaining the zero bit number information from the bit string corresponding to the maximum possible value; An address information obtaining unit for obtaining address information of the decoding table using at least one of the zero bit number information and the remaining bit information indicating the remaining bits present after the first bit 1 of the codewords; And a total zero information selector configured to select total zero information from the decryption table using the address information.

The apparatus for restoring total zero information further includes a memory configured to store a new decryption table generated by using the characteristics of the codewords, and the upper four bits of total zero information and a lower order from a memory storing the decryption table by using the address information. The present invention provides a total zero information restoring apparatus based on a decoding table characterized by reading 4-bit total zero information.

The total zero decoding apparatus further includes a 4-bit comparison unit configured to compare upper 4-bit total zero information and lower 4-bit total zero information read from a memory storing the decoding table, and the upper 4-bit total zero information and the lower 4 bits. If the total zero information is not the same, after one additional bit is read from the input bit string buffer, one of the upper four bits total zero information and the lower four bits total zero information is selected, and the upper four bits total zero information and the lower four are selected. When the bit total zero information is the same, the upper 4 bit total zero information is selected.

The total zero recovery apparatus further includes an input bit string calculator that obtains bit string length information using at least one of the zero bit number information, and corrects a pointer of an input bit string buffer using the bit string length information. It features.

The method and apparatus for restoring total zero information using the new decryption table according to the present invention, unlike the conventional scheme, provide an efficient memory structure for storing total zero information, thereby reducing repetitive memory access.

Specifically, the method and apparatus for restoring total zero information by using the new decryption table according to the present invention can reduce the power consumption of the system by reducing the memory access by about 80-90% compared to the conventional method, and at the same time increase the operating speed of the system. There is an increasing effect. In addition, the memory structure proposed in the present invention has an effect of not causing an unnecessary increase in memory.

The terminology used in the present invention is a general term that is currently widely used as possible, but in certain cases, the term is arbitrarily selected by the applicant, in which case the meaning of the term is described in the detailed description of the invention. It should be understood that the present invention in terms of terms other than these terms.

Hereinafter, with reference to the accompanying drawings a preferred embodiment of the present invention that can specifically realize the above object, the present invention is not limited or limited by the above embodiments.

2 is a table reconstructing the table of FIG. 1 according to a non-zero coefficient number value Tc according to an embodiment of the present invention. Although the present invention is described using the VLCT used in H.264 / AVC as an example, the present invention is not limited to H.264 / AVC, but may be applied to all decompression apparatuses and methods for decoding total zero information.

In order to recover total zero information at high speed, i) memory access should be replaced by arithmetic operations as much as possible, and ii) new decryption tables that can reduce memory access when it is impossible to replace arithmetic operations, and memory structures that efficiently store them. Should be designed.

In order to solve the above requirements, it is important to understand the codeword characteristics of FIG. 1. To this end, the present invention analyzes and reconstructs FIG. 1 according to the number information Num_Z of the bit 0 existing before the first bit 1. FIG. 2 shows an example of reconstruction for Tc = 1 to Tc = 8. 1 and 2, the following features can be seen.

1) Since total zero information is 0 ~ 15, it can be represented by 4 bits.

2) Up to four codewords having the same number of zero bits information (Num_Z) exist.

3) The remaining bit information (iNFO) present after the first bit 1 is a maximum of 2 bits.

4) Codewords having the same number of zero bits information Num_Z are accurately classified into the remaining bit information iNFO.

5) For Tc = 1,14,15, the relationship between total zero information and codeword is simple.

Based on these characteristics, the present invention divides an existing restoration method into a restoration method based on an arithmetic operation and a restoration method based on a new decoding table. Hereinafter, a restoration method based on an arithmetic operation and a restoration method based on a new decoding table will be described in detail.

1. Arithmetic-based Restoration

In the case of Tc = 1, the total zero information may be arithmetic calculated as in Equation 1 using the zero bit number information and the remaining bit information.

Referring to FIG. 2, when the codeword CW is 1, since the zero bit number information is zero, the total zero information is zero. When the codeword is 00011, since zero bit number information is 3 and the remaining bit information is 1, the total zero information is 3 * 2-1 = 5. When the code is 000000001, since the zero bit number information is 8, the total zero information is 8 * 2-1 = 15.

In addition, when Tc = 14, 15, the total zero information may be arithmetic calculated as in Equation 2 using the input bit string information.

Here, IN [1: 0] represents input bit string information composed of two bits. Referring to FIG. 1, when Tc is 15 and the codeword is 0, the total zero information is 0 because IN [1] = 0, and when Tc is 15 and the codeword is 1, the total zero information is 0. The information is one. If Tc is 14 and the codeword is 01, then IN [1] = 0, IN [0] = 1, so the total zero information is 0 * 2 + 1 * (1-0) = 1, and Tc is 14 and the codeword is In the case of 1, since IN [1] = 1, the total zero information is 1 * 2 + IN [0] * (1-1) = 2.

In Equations 1 and 2, however, the length of the input bit string information used for calculating the total zero information is not fixed. In other words, the length of the bit string used varies. Therefore, in order for the correct bit string to be used for Run_Before restoration performed after total zero information restoration, the pointer of the input bit string buffer needs to be corrected. The length information bit_rd_len of the bit string read from the input bit string buffer in the calculation of Equations 1 and 2 can be obtained using Equation 3 below.

When Tc = 15, the bit string length information is always one. When Tc = 14, the bit string length information is 1 only when the input bit string information is 1, that is, the code word is 1. In the case of other code words, the bit string length information is 2. When Tc = 1 and zero bit number information is 0, the bit string length information is 0 + 1 = 1. If Tc = 1 and the zero bit number information is 8, the bit string length information is 8 + 1 = 9. In other cases, bit string length information is obtained by adding 2 to the zero bit number information of each codeword.

The reconstruction method based on arithmetic operations from Equations 1, 2, and 3 is performed by the following process.

First, if Tc = 1, the zero bit number information is obtained and 1 bit Read the remaining bit information. If Tc = 14,15, the input bit string information IN [1: 0] is read.

Second, total zero information is calculated according to equations (1) and (2).

Third, the bit string length information is obtained according to Equation 3 to correct the point of the input bit string buffer.

As described above, if the total zero information cannot be restored through an arithmetic operation, the total zero information is restored using the new decoding table as follows.

2 new decrypted table-based restore

In the conventional reconstruction scheme, the VLCT of FIG. 1 is implemented by storing a corresponding codeword in a memory using total zero information and address information generated by a Tc value. Based on this, the codeword matching the input bit string is obtained through repetitive memory access, and then the total zero information is restored from the obtained address information.

However, the codeword features of 1) to 4) described above provide an idea of a new decoding table generation and its memory structure design. First, since there are at most four codewords having the same number of zero bits, address information may be generated using Tc and zero bits. In addition, two total zero information is represented by 8 bits. Therefore, two total zero information having the same number of zero bits can be stored in the positions of the upper 4 bits (U4) and the lower 4 bits (L4) according to the LSB of the codeword. From this fact, the present invention proposes a memory structure for storing a new decryption table as shown in FIG.

3 is a memory structure for storing a new decoding table generated by using characteristics of codewords according to an embodiment of the present invention. The new decoding table does not include the existing codeword but includes total zero information. The underlined Add value in FIG. 3 is a base address (B_add), which is determined by Equation 4 below. Offset is an offset address for codewords belonging to a specific Tc value and is calculated as in Equation 5 below. Therefore, Add in FIG. 3 is obtained by the sum of B_add and offset. In addition, underlined redundant memory and 2 bytes of unused memory are intentionally inserted for the convenience of Add calculation. Even if such surplus memory is inserted, the amount of memory of the proposed memory structure is smaller than that of the existing memory structure.

Here, i, j, k, m are control information generated by each Tc. Hereinafter, a method of restoring total zero information by using a memory for storing the new decoding table will be described in detail.

4 is a flowchart illustrating a method of restoring total zero information at high speed according to an embodiment of the present invention. Referring to FIG. 4, it is first determined whether total zero information can be calculated from a conventional VLCT through an arithmetic operation (S401). If total zero information can be calculated through an arithmetic operation, necessary information is obtained according to a Tc value among zero bit number information, remaining bit information, and additional input bit string information (S402). Next, total zero information is calculated using Equation 1 or Equation 2 (S403). Next, bit string length information is obtained using Equation 3, and the pointer of the input bit string buffer is corrected using the bit string length information (S404).

If the total zero information cannot be calculated through an arithmetic operation, the total zero information is restored as follows based on the memory structure of the new decoding table according to the present invention.

First, for a given Tc Control information (i, j, k, m) is obtained (S405).

Second, the maximum possible value Max_Z_len of the zero bit number information is obtained using Equation 6 below (S406).

Third, zero bit number information is obtained from a bit string corresponding to the maximum possible value (S407).

Fourth, address information Add is obtained using Equations 4 and 5 (S408).

Fifth, the upper 4 bits (U4) and the lower 4 bits (L4) are read from the memory for storing a new decoding table (S409).

Sixth, if U4 ≠ L4, one bit is additionally read from the input bit string buffer, and one of U4 and L4 is selected, and if U4 = L4, U4 is selected (410).

Seventh, the bit string length information is generated as shown in Equation 7 below to correct the pointer of the input bit string buffer (404).

Since the method for obtaining bit string length information using Equation 7 is similar to the method for obtaining bit string length information with reference to Equation 3, a detailed description thereof will be omitted.

5 illustrates an apparatus for restoring total zero information at high speed according to an embodiment of the present invention. The decompression device may be used to decompress total zero information using a new decoding table. Referring to FIG. 5, the decompression device includes a control information and a maximum possible value generator 502, an address information calculator 503, a zero bit number information acquirer 504, a decoding table storage memory 506, and a bit string. A length calculator 507, a 4-bit comparison unit 508, and a total zero information selector 510 are included.

The control information and maximum possible value generation unit 502 generates the control information (i, j, k, m) using the Tc value input from the Coeff_token decompression device, and using the equation (6) of the zero bit number information Generate the maximum possible value.

The zero bit number information obtaining unit 504 obtains the number of '0's existing before the first bit 1 from the bit string stored in the input bit string buffer 505. In this case, the maximum possible value may be used.

The address information calculator 503 calculates the address information by using the control information generated by the control information and the maximum possible value generator 502 and equations (4) and (5). In this case, input bit string information may be used from the input bit string buffer 505 if necessary. The upper four bits U4 and the lower four bits L4 are read from the decoding table storage memory 506 using the calculated address information.

The total zero information selector 510 selects one of the U4 and L4. If U4 and L4 are different, the total zero information selector 510 selects one of U4 and L4 after additionally reading one bit from the input bit string buffer 505. If U4 and L4 are the same, select U4 or L4.

The bit string length calculator 507 calculates bit string length information by using the control information, zero bit number information, and Equation 7, and transmits the bit string length information to the input bit string buffer 505 to input the bit string length information. Correct the pointer to the bit string buffer.

6A and 6B are graphs showing memory access gains in a total zero information restoration method based on a new decoding table according to an embodiment of the present invention. In the present invention, the performance of the total zero reconstruction method according to the present invention was verified using QCIF (Quarter CIF) and CIF (Comm on Intermediate Format) test images. At this time, the search range is set to 16 and 5 prediction frames are applied. The R-D optimization technique was applied to test image compression. Other experimental environments were based on the conditions recommended in the H.264 / AVC Baseline profile.

6A and 6B, the verification was performed on foreman, news, silent, and conjugate images, and it can be seen that the memory access ratio is significantly reduced compared to the conventional method. 6A and 6B, MAG is defined as in Equation 8 below.

In the present invention as described above has been described by the specific embodiments, such as specific components and limited embodiments and drawings, but this is provided to help a more general understanding of the present invention, the present invention is not limited to the above embodiments. For those skilled in the art, various modifications and variations are possible from such description.

Therefore, the spirit of the present invention should not be limited to the described embodiments, and all the things that are equivalent to or equivalent to the claims as well as the following claims will belong to the scope of the present invention. .

1 is a variable length codeword table for a conventional total zero syntax element.

2 is a table reconstructing the table of FIG. 1 according to a non-zero coefficient number value according to an embodiment of the present invention. FIG.

FIG. 3 is a memory structure for storing a decoding table generated using characteristics of codewords according to an embodiment of the present invention, in which a total zero value to be finally restored is stored in upper and lower 4 bits.

4 is a flowchart illustrating a method of decoding total zero information at high speed according to an embodiment of the present invention.

5 is an apparatus for decoding total zero information at high speed according to an embodiment of the present invention.

6A and 6B are graphs showing memory access gains in a total zero information restoration method based on a new decoding table according to an embodiment of the present invention.

<Description of the symbols for the main parts of the drawings>

502. Control information and maximum possible value generator

503. Address information calculation unit

504. Zero bit number information acquisition unit

505. Input Bit String Buffer

506. Decryption table storage memory

507. Input bit string information calculation unit

508. 4-bit comparison unit

510. Total Zero Information Selector

Claims (14)

A method of decoding total zero information (Total_Zeros) based on a decoding table in a decoding apparatus, Obtaining non-zero coefficient number information Tc indicative of the number of non-zero coefficients; Calculating the total zero information through an arithmetic operation when the relationship between the code words corresponding to the non-zero coefficient number information and the total zero information is simple; And Acquiring the total zero information by using a new decoding table generated by using characteristics of the code words when the relationship between the code words corresponding to the non-zero coefficient number information and the total zero information is not simple; A low power total zero information restoration method based on a new decoding table comprising a. The method of claim 1, wherein the step of obtaining total zero information Generating control information (i, j, k, m) using the non-zero coefficient number information; Obtaining a maximum possible value of the zero bit number information using the control information; Obtaining the zero bit number information from the bit string by the maximum possible value; Calculating address information (Add information) of the new decoding table using the zero bit number information and the remaining bit information (iNFO) representing the remaining bits existing after the first bit 1 of the codewords; And And selecting total zero information from the new decryption table by using the address information. The method of claim 2, The new decoding table does not include the existing codeword, and includes the total zero information, low power total zero information recovery method based on the new decoding table. The method of claim 2, wherein the step of selecting total zero information Reading upper 4-bit total zero information (U4) and lower 4-bit total zero information (L4) from a memory storing the new decoding table using the address information; And If the upper 4 bits total zero information and the lower 4 bits total zero information are not the same, after additionally reading 1 bit from the input bit string buffer, the upper 4 bits total zero information U4 and the lower 4 bits total zero information L4 And selecting the upper 4 bits total zero information or the lower 4 bits total zero information when the upper 4 bits total zero information and the lower 4 bits total zero information are the same. Low power total zero information restoration method based on new decoding table. The method of claim 2, The control information is defined as in Equation 1 below, and the address information is obtained from a sum of B_add and offset values obtained through Equation 1 and Equation 2 below. Based on low power total zero information recovery method. <Equation 1>

<Equation 2>

The method of claim 2, The maximum possible value is a low power total zero information recovery method based on a new decoding table, characterized in that obtained by using the following equation (3). <Equation 3>

The method of claim 2, wherein the obtaining of the total zero information is performed. Obtaining bit string length information using the zero bit number information, and then correcting a pointer of an input bit string buffer using the bit string length information, wherein the low power is based on a new decoding table. How to restore total zero information. The method of claim 7, wherein The bit string length is a low power total zero information recovery method based on a new decoding table, characterized in that obtained using Equation 4. <Equation 4>

In the low power total zero information recovery apparatus based on a decoding table, Control information (i, j, k, m) is generated using the non-zero coefficient number information, and control information and maximum possible value generation are generated using the control information to generate the maximum possible value of the zero bit number information. part; A zero bit number information obtaining unit for obtaining the zero bit number information from the bit string corresponding to the maximum possible value; An address information calculator configured to calculate address information of a new decoding table using the zero bit number information and the remaining bit information indicating the remaining bits present after the first bit 1 of the codewords; And And total zero information selection unit for selecting total zero information from the new decryption table by using the address information. 10. The apparatus of claim 9, wherein the total zero information decompression device is Further comprising a memory for storing a new decoding table created by using the characteristics of the code words, An apparatus for recovering low power total zero information based on a new decryption table according to claim 4, wherein upper 4-bit total zero information and lower 4-bit total zero information are read from a memory storing the decryption table using the address information. The method of claim 10, And a memory structure for storing the new decryption table has a structure as shown in FIG. 3. The apparatus of claim 11, wherein the total zero recovery device is And a 4 bit comparison unit for comparing the upper 4 bits total zero information and the lower 4 bits total zero information read from the memory storing the decoding table. If the upper 4 bits total zero information and the lower 4 bits total zero information are not the same, after one additional bit is read from an input bit string buffer, one of the upper 4 bits total zero information and the lower 4 bits total zero information is selected. And when the upper 4 bits total zero information and the lower 4 bits total zero information are the same, the upper 4 bits total zero information or the lower 4 bits total zero information is selected. Information Restoration Device. The apparatus of claim 9, wherein the total zero recovery device is And a bit string length calculator which obtains bit string length information using the zero bit number information. And a new decoding table based on the new decoding table, wherein the pointer of the input bit string buffer is corrected using the bit string length information. A computer-readable recording medium having recorded thereon a program for executing the method according to any one of claims 1 to 8.

Images