KR20000021048A - Variable length coding table structure and method for coding variable length using the same - Google Patents

Abstract

PURPOSE: A method for coding variable length using a variable length coding table is provided to process coding of a variable length coder of an MPEG2 encoder at a high speed in real time by reading and outputting a variable length code. CONSTITUTION: A rearranged variable length coding table enables variable length code values corresponding to index values to be instantly outputted through a conversion of quantized input data values into indexing values(500). A final output is a code and a length. A variable length code is the same value, but the length is different, so the variable length code is not indicated with only the value. Therefore, the length of the code is outputted together. Input data to be coded are values which are sequentially stored in a memory and outputted from the memory.

Description

Variable length coding table structure and variable length coding method using the same

The present invention relates to a real-time processing method of a variable length coder (Variable Length Coder) in the MPEG2 encoder (encoder), in particular to rearrange the table of the variable length coder and variable length encoding the input data using the rearranged table, The present invention relates to a structure of a variable length coding table that enables high-speed data processing by executing a sequential scan in a conventional zigzag scan and a variable length coding method using the same.

There are many ways to implement an MPEG-2 encoder, but one of the typical methods is to perform ASIC using hardwire logic.

The advantages of this method are that it can be designed compactly for the purpose of use, and the number of gates is lower than other methods, but it is difficult and expensive to correct when the ASIC design or specification is wrong.

Other methods of implementing the MPEG2 coder are those that require a lot of computation time, such as DCT or motion estimation, which are implemented in hardwired logic, and the rest are designed with a digital signal processor. It is easy to fix when the ASIC design or specification goes wrong, it is very inexpensive, and it is easy to upgrade the IC next time.However, because the digital signal processor is general purpose, the number of gates and the real-time processing are difficult. There are disadvantages.

1 is a block diagram of a conventional MPEG2 encoder, which includes a DCT unit 101 for DCT transforming input digital image data, a quantizer 102 for quantizing DCT transformed coefficient values, and variable length encoding of quantized data. A variable length encoder 103, a buffer 104 for outputting a variable length coded image data stream, a bit rate controller 105 for adjusting a bit rate for the quantization, and the quantized data An inverse quantization unit 106 for inverse quantization, an inverse DCT unit 107 for inverse DCT processing the inverse quantized data, a frame memory 108 for storing the inverse DCT processed data, and the stored data And a motion estimator 109 and a motion compensator 110 for respectively performing motion estimation and compensation.

The DCT unit 101 performs a discrete cosine transform (DCT) on the difference between the digital video input signal and the output of the motion estimation unit 109 of the previous video signal. Is quantized under the control of the bit rate control unit 105 and input to the variable length encoder 103.

The variable length encoder 103 uses Huffman Coding and Run Length Coding, which are noiseless coding for approaching the entropy of source data, to form the quantized DCT coefficient ( The coefficients are compressed and output as a video data stream (MPEG2 Video Bitsream) through the buffer 104.

Meanwhile, the quantized signal is inversely quantized by the inverse quantization unit 107, inversely transformed by the inverse DCT unit 107, and stored in the frame memory 108.

A motion vector is generated using data stored in the frame memory 108 and input digital image data, and motion compensation is performed by the motion compensator 110 using the motion compensation data. The difference between the two is obtained and input to the DCT unit 101.

This series of operations is performed to MPEG-2 encode the input digital image data. The biggest problem in implementing a variable length encoder using a digital signal processor is whether the data can be processed in real time.

The variable length coding is to map a DCT coefficient to a code word previously prepared using Huffman coding and run length coding in MPEG2.

However, when the input video data format is 4: 2: 0, the number of data to be processed by the variable length encoder is 1.5 times the number of pixels of the image data, for example, the resolution of the image data ( If the resolution is 720 × 480, then 720 × 480 × 1.5 = 518440.

In case of NTSC signal, variable length encoder must process 518440 data for 33.33msec.

Therefore, a processing time of about 64 nsec is given to one pixel, which is a very short time for the digital signal processor to process, making it difficult to process in real time and causing bottlenecks at the input / output stages of data, thereby degrading system performance. do.

That is, if the variable length encoder is implemented using the conventional MPEG2 variable length coding table as it is, the coding of the digital signal processor becomes complicated and the execution speed is also lowered.

On the other hand, in the MPEG2 standard, the order of processing block data in a macro block (in MPEG, the data is organized in a hierarchical structure, one macro block is composed of four blocks) is shown in FIG. 2A. It is specified in order or alternative order.

That is, since the data is crowded into the DC component after the DCT, variable coding in the zigzag scanning order can compress a larger amount of data.

However, zigzag scanning introduces many problems when performing variable length coding using a digital signal processor.

That is, after performing DCT, the data length is 16 bits, but the digital signal processor should be able to read a large amount of data at once to speed up the execution. However, if the data is stored by zigzag scanning as shown in FIG. High speed real time processing was difficult.

The present invention provides a variable length encoding table structure that enables real-time high-speed processing when performing variable length encoding of an MPEG2 encoder using a digital signal processor, and a variable length encoding method using the same.

The present invention has a new relocated variable length encoding table structure in which an existing variable length encoding table is indexed by arranging an existing variable length encoding table from '0' to a constant value according to the order of input codes. The present invention provides a method for enabling real-time variable length encoding processing by encoding by means of encoding.

In addition, the present invention, after performing the DCT in the DCT unit can be stored in a sequential order to store the output data in the block unit in a zigzag scanning order to read and process a large amount of data in a bundle at a time A variable length coding table structure and a variable length coding method using the same are provided.

1 is a block diagram of an MPEG2 encoder

2A is a diagram illustrating a zigzag scan in a conventional encoder.

2b illustrates a sequential scan in the encoder of the present invention.

3 is a block diagram of a variable length coder according to the present invention.

4 is a diagram illustrating an example of a variable length encoder table structure to which the present invention is applied.

5 is a diagram illustrating an example of a structure of a variable length encoder table rearranged according to the present invention.

3 is a block diagram of a variable length encoder according to the present invention. The DCT coefficients quantized by the DCT and the quantization unit 301 are transformed and stored in the order of the zigzag scanning by the zigzag scanning conversion unit 302 and rearranged them. The variable length encoder 303 including the variable length coding table encodes.

FIG. 4 illustrates a variable length coding table for B picture as an example of a conventional variable length coding table for explaining an example of a rearranged variable length coding table of the present invention, and FIG. 5 shows a variable length coding table for B picture of FIG. The variable length coding table of the present invention configured by rearranging the length coding table is shown.

The construction method of the variable length coding table and the variable length coding method using the same will be described as follows.

If the structure of the variable-length coding table for the DCT and quantized input data as shown in FIG. 3 is the same as that of FIG. 4, the input codeword 400 is indexed to the codeword value and the index value as shown in FIG. Rearrange them to 500 and map the corresponding variable length code values (VLC codes) to arrange the variable length coding table from '0' to a constant value ('0' to '31' in FIG. 5). Make sure there are no missing values (ie make VLC code = 0).

As shown in FIG. 4, the variable length code of the Intra type macro block 401 is '00011', and when the values of the respective flags 400 are combined, the value becomes '1' and is arranged at the indexed value 501 = 1 as shown in FIG. 5. In FIG. 4, the variable length code of the Intra, Quant type macro block 402 is '000001', and the sum of the values of the respective flags 400 becomes '17', thereby indexing the value 502 = 17 as shown in FIG. Will be arranged at.

When indexing is performed on the remaining data in this manner, the existing variable length encoding table of FIG. 4 is rearranged as shown in FIG.

As shown in FIG. 5, the rearranged variable length coding table immediately undergoes a step of converting the DCT and the quantized input data value 400 into the indexing value 500, and then the variable length code value (VLC code) corresponding to the corresponding index value. To print immediately.

In this case, the final output comes out as a code and a length. This is because variable length codes have the same value but usually have different lengths, so the length of the code must be output as well.

On the other hand, the input data to be encoded by the rearranged variable-length encoding table are values that are sequentially stored in the memory in a zigzag scanning order and output as described above.

That is, in the scanning conversion unit 302 of FIG. 3, when data is stored after DCT, the data is sequentially stored as shown in FIG. 2B without storing the data in the zigzag scanning order as shown in FIG. 2A.

In this case, the variable length encoder 303 having the rearranged encoding table of the present invention can read several data at the same time, thereby increasing the execution speed.

Of course, the order of data to be written to the block is written sequentially as shown in FIG. 2B, but the actual content is maintained in the zigzag order of FIG. 2A. Therefore, if the data of the block is read sequentially from the memory, the data is read in zigzag order. Being the same, there is no problem in signal processing.

The present invention includes a rearranged variable length coding table in which a variable length code is mapped to an existing variable length coding table indexed and arranged in order. The present invention converts an input value into an index value and converts the variable to an index value. By allowing the length code to be immediately read and output, the variable length encoder of the MPEG2 encoder can perform encoding in real time and high speed.

In the present invention, the variable length encoder can store the data read out in the zigzag scanning order sequentially in the memory and read out several data at a time, thereby enabling high-speed real-time processing of the variable length coding of the MPEG2 encoder.

Claims (4)

In compression encoding by receiving data in block units and performing data transform and quantization, Variable length coding table having a structure in which input lengths are indexed and sequentially arranged in a corresponding structure of a variable length coding table, and the rearranged index values are mapped to variable length code values. rescue. The intermediate value of claim 1, wherein the index value is a value from '0' to a predetermined level in sequence, and when the variable length code value is not present in the rearranged index value, the index value is assigned as a code value. A variable length encoding table structure, wherein the structure is such that there is no value missing. In the method of compression encoding by receiving data in block units and performing data transformation and quantization, Variable length code value of variable length coding table: The variable length coding table of the structure that indexes input values in the corresponding structure of input values and arranges them sequentially, and rearranges this rearranged index value: variable length code value, input value And converting the variable length into an index value and reading and outputting a variable length code value corresponding to the converted index value from the rearranged variable length coding table. The variable length encoding method of claim 3, wherein a value input for the variable length encoding is a value stored sequentially in a memory while having a zigzag scanning order.