KR20040067239A - Method of bit allocation in quantizer, and apparatus thereof - Google Patents

Abstract

PURPOSE: A method for assigning bits used for performing quantization is provided to linearly approximate the number of assigned bits for quantization, in order to effectively carry out the quantization, thereby maintaining high accuracy even though a scene change is generated. CONSTITUTION: A system receives DCT coefficients for one macro block of one inputted image frame(S410). The system counts the number of '0' among the DCT coefficients within the inputted macro block, and calculates the ratio of the counted number of '0'(S420). The system decides whether the ratio of the counted number of '0' is calculated for all macro blocks of the one image frame(S430). If some macro blocks to be inputted remain, the system receives a next macro block, and receives all the macro blocks. If the calculation of the ratio of the counted number of '0' is completed, the system assigns predetermined bits by linearizing with the use of the calculated ratio of the counted number of '0' (S440).

Description

Method and apparatus for allocating bits used when performing quantization {Method of bit allocation in quantizer, and apparatus}

The present invention relates to video coding, and more particularly, to a method and apparatus for adjusting the number of bits allocated when performing quantization in video coding using a method such as MPEG.

The video compression method proposed by Moving Picture Expert Group (MPEG) is based on the method of eliminating temporal redundancy and spatial redundancy. That is, temporal redundancy is removed by using the moving compensation compression method, and spatial coherence is removed by applying DCT (Discrete Cosine Transform) to still images.

FIG. 1 is a diagram illustrating a video compression process of MPEG.

That is, after receiving the original image (S110), performing the preprocessing process (S120) and converting to the frequency domain (domain) (S130). In the preprocessing step S120, color space conversion, filtering, and color subsampling may be performed as a preparation step for performing compression. In the conversion step (S130), the input image is converted from a spatial domain to another region by using various types of mathematical transformation methods in order to remove redundancy of the input image data. DCT is used during this conversion.

When DCT is performed, DCT coefficients are generated, and quantization is performed using the generated DCT coefficients (S140). Then, the high frequency component is removed and only the low frequency component remains, so only the low frequency component is encoded (S150). Therefore, the compression efficiency can be improved. The generated coded bit stream is output (S160). In the converting step S130, the DCT is performed in macro block units. When DCT is performed on an 8 × 8 macroblock, a matrix A consisting of DCT coefficients is generated. In the matrix A generated as described above, low frequency components and high frequency components exist together. In general, important data are mainly located in low frequency components, and thus high frequency components are mainly removed in the quantization process S140.

As coding, variable length coding (VLC) is mainly used. Variable Length Coding (VLC) is a method of reducing the size of an entire file by representing data that appears frequently according to the frequency of appearance of data in a small number of bits and data that does not appear frequently in a relatively large number of bits. Huffman coding is used as the processing method according to the frequency.

In MPEG-4, which is an object based coding method, a method of using a fixed quantization bit in providing a variable bit rate (VBR) service is still widely used. When using this fixed quantization bit, if the amount of data in the output bit stream changes too much, there is a high possibility that an overflow or underflow of the output buffer occurs. In addition, in constant bit rate (CBR) service, it is necessary to flexibly adjust the quantization bit so that the amount of output data is constant regardless of the type of the input image. This is called rate control. do.

The bit rate control method includes a frame-based bit rate control method for adjusting the amount of output data in units of frames and a macro block based bit rate control method for adjusting the amount of output data in units of macroblocks in a frame. Although the macroblock-based bitrate control method can more accurately control the bitrate, the frame-based bitrate control method is generally used because the control method is too complicated and difficult.

In the frame-based bit rate control algorithm in MPEG-4 VM18 (Verification Model 18), the Quadratic Rate Distortion (RD) model is expressed by Equation 1 below.

Here, T means target bits, that is, the number of bits used in quantization, and MAD means the sum of the absolute values of the difference between the input image and the motion compensated reconstructed image. In addition, QP represents a quantizer, and X ₁ and X ₂ are rate distortion (RD) modeling parameters, which are values updated after an encoding process for each input image frame is performed.

2A is a diagram illustrating Equation 1 as a graph.

Referring to FIG. 2A, it can be seen that the nonlinear characteristic is shown. In this way, it is necessary to change the nonlinear characteristic linearly, and even though Equation 1 is scalable, the DCT coefficient is equal to every input frame. Since the calculation process for controlling the bit rate is complicated, especially when the equation 1 is applied to the quantization as it is when there is a scene change, the accuracy of the model is poor.

The technical problem to be solved by the present invention is to allocate the bits used in the quantization to efficiently perform quantization by linearly approximating the number of bits allocated in the quantization used in the video compression method as in MPEG 4. To provide a method and an allocation device.

FIG. 1 is a diagram illustrating a video compression process of MPEG.

2A is a diagram illustrating Equation 1 as a graph.

2B is a characteristic graph when the bit allocation method of the present invention is used.

3A to 3B are diagrams for explaining a process of calculating a coefficient ρ used to determine bits allocated in quantization.

4 is a flowchart of the bit allocation method of the present invention described above.

5 is a block diagram of a bit allocation apparatus to which the bit allocation method of the present invention described above is applied.

6 is a graph showing the relationship between the number of bits used and the number of bits actually used when the bit allocation method of the present invention is used.

7 is a graph of experimental results when an image titled "mobile" having a QCIF size is input.

In order to achieve the above object, a bit allocation method used in performing quantization according to the present invention includes the steps of: counting the number of zeros present in a DCT coefficient for a macroblock; And reducing the number of bits to be allocated during quantization if the number of counted zeros is large, and increasing the number of bits to be allocated during quantization if the number of zeros counted is small.

In order to achieve the above object, a bit allocation method used in performing quantization according to the present invention includes: receiving a DCT coefficient for a macro block; Counting the number of zeros among the DCT coefficients in the input macroblock and calculating a ratio p (i, j) of the number of zeros; Determining whether a value of ρ (i, j) is calculated for all macroblocks of one image frame; And if the macroblock to be input remains as a result of the determination and receives the next macroblock, and if all the macroblocks are received, performing a predetermined bit allocation by linearizing using the value of ρ (i, j). do.

In order to achieve the above object, an apparatus for allocating bits used in quantization according to the present invention includes: an image input unit sequentially inputting DCT coefficients for one macro block of one image frame in macroblock units; A counting unit for counting a number of zeros of DCT coefficients in the input macro block; An intermediate value calculator for calculating a ratio p (i, j) of the number of zeros counted by the counter; And an allocation bit determiner configured to linearize using the value of ρ (i, j) calculated by the intermediate value calculator to determine a bit to be used for quantization.

In order to achieve the above object, the present invention provides a computer-readable recording medium recording a program for executing the method on a computer.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

2B is a characteristic graph when the bit allocation method of the present invention is used.

When the bit allocation method using Equation 1 described above is used, a nonlinear characteristic is obtained, whereas when the bit allocation method of the present invention is used, a linear characteristic is obtained. The linear characteristics will be described later.

Assume that video stream S is encoded with an MPEG-4 encoder with a fixed bit quantizer j, and express the encoded bit stream for frame i as T (i, j). If the video stream S is encoded by an MPEG-4 encoder having a fixed bit quantizer k, the encoded bit stream for the frame i may be represented by T (i, k).

Then, there is a linear relationship between the encoded bit streams T (i, j) and T (i, k) as shown in Equation 2 below.

Where j and k denote quantizers as described above, T (i, j) and T (i, k) denote encoded bit streams for frame i, and N denotes the total number of frames of the input video stream S Indicates.

Quantization is performed after Motion Compensation (MC) and DCT are performed in MPEG-4 video compression. DCT coefficients are obtained by quantizer j. The quantized coefficients are rearranged into a 1-D array L by a zigzag scan in each macro block.

Run length is counted for each successive zero between two nonzero coefficients in a reordered one-dimensional array L. For example, if a zigzag scan yields L as "5 8 0 4 0 0 0 8 0 0 ... 0 2 0 0 ... 0", then a non-zero constant (the Length, the value of a non-zero constant) This reduces the overall amount of data. Then we calculate the sum of the sizes of all run lengths and express this value as Q ' _z . Here, the nonzero number size means the number of bits when expressed as sign magnitude.

For example, if +17 is represented as a binary number, since it is "10000", the size may be referred to as 6 bits by adding 1 bit of a sign bit and 5 bits indicating a magnitude. In this way, the sum of the sizes of all nonzero coefficients can be calculated.

If the calculated frame size is M, then Q _nz and Q _z can be defined for frame i as follows. And p (i, j) represents the ratio of zero in the quantized DCT coefficients in frame i.

3A to 3B are diagrams for explaining a process of calculating a coefficient ρ used to determine bits allocated in quantization.

A method of calculating ρ (i, j) will be described in detail with reference to FIG. 3. It is assumed that an image of a Quadrature Common Intermediated Format (QCIF) is input as shown in FIG. 3A. Since QCIF is 1/4 the size of CIF, the number of pixels in one frame is 176x144. Assuming that a DCT coefficient as shown in FIG. 3B is obtained as a result of performing DCT on the first macroblock 1 310, p for the first macroblock 1 310 is calculated by the following equation (4). For example, if the number of zeros in A ₁ is five, then 5 / (176x144).

Calculated as in the following as the macro block 2 was carried out the DCT about 320 to obtain a DCT coefficient A _2, if the number of 0 to 10 among the DCT coefficients present in manil _{A 2 (5 + 10) /} (176x144) do. Thus, the result of accumulating the number of zeros in the DCT coefficients for all macroblocks is ρ (i, j). The number of quantization bits is determined using the coefficient p (i, j) thus obtained. In other words, if the value of ρ (i, j) is small, the number of bits allocated for quantization is increased. If the value of ρ (i, j) is large, the number of bits allocated for quantization is reduced.

The smaller the value of ρ (i, j) means that there are fewer zeros in the DCT coefficients, and the larger the value of ρ (i, j) means that the greater the number of zeros in the DCT coefficients. Since it means that there are few components to be encoded, the bits used in quantization are adaptively adjusted according to the amount of data to be encoded.

Then, θ (i, j) is calculated by the following Equation 5 using the calculated value of ρ (i, j), and finally the final allocation bit T (i for the frame i encoded with the quantizer j , j) is calculated by the equation (6).

T (i, j) = k (i, j) * rho (i, j) + c (i, j)

Where k (j, j) and c (i, j) are given by Equations 7, 8 and 9 according to the range of j values. In other words, in the case of 1 ≦ j ≦ 5, equation 7 is applied, in the case of 6 ≦ j ≦ 18, equation 8 is applied, and in case of 19 ≦ j ≦ 31, equation 9 is applied to k (j, j) and c. (i, j) is calculated.

4 is a flowchart of the bit allocation method of the present invention described above.

In operation S410, a DCT coefficient of one macroblock of an input image frame is received. Then, the number of 0s in the DCT coefficients in the input macro block is counted to calculate the value of ρ (i, j) (S420). The value of ρ (i, j) is calculated by the above equation (4). It is determined whether the value of ρ (i, j) has been calculated for all the macroblocks of one image frame (S430), and there is still a macroblock to be input, the next macroblock is received, and all the macroblocks are received and the ρ (i , j) When the calculation of the value is completed, linearization is performed using the calculated value of ρ (i, j) to perform a predetermined bit allocation (S440). If the value of ρ (i, j) is large, the bit allocation is small. If the value of ρ (i, j) is small, the bit allocation is large.

Summarizing the bit allocation method described in FIG. 4, the number of zeros present in the DCT coefficients for the macroblock is counted. If the number of counted zeros is large, the number of bits to be allocated during quantization is reduced, and the number of counted zeros is counted. It can be seen that less is to increase the number of bits to allocate during quantization.

5 is a block diagram of a bit allocation apparatus to which the bit allocation method of the present invention described above is applied.

The image input unit 510 sequentially receives DCT coefficients of one macro block of one image frame in macroblock units. The counting unit 520 counts the number of zeros of the DCT coefficients in the input macro block. The median value calculator 530 calculates a value of ρ (i, j) using Equation 4 described above. The allocation bit determiner 540 linearizes the value of ρ (i, j) calculated by the intermediate value calculator 530 to determine a bit to be used during quantization.

6 is a graph showing the relationship between the number of bits calculated when the bit allocation method of the present invention is used and the number of bits actually used.

The image used in the experiment is an image titled "foreman" of 100 frames of QCIF size. Referring to FIG. 6, it can be seen that the number of bits in the case of using the bit allocation method of the present invention is almost identical to the number of bits actually used. 6 denotes the number of bits actually used, and denoted by * denotes the calculated number of bits when the bit allocation method of the present invention is used.

7 is a graph of experimental results when an image titled "mobile" having a QCIF size is input. Referring to FIG. 7, it can be seen that similar results can be obtained when an experiment is performed by inputting another test image.

The invention can also be embodied as computer readable code on a computer readable recording medium. The computer-readable recording medium includes all kinds of recording devices in which data that can be read by a computer system is stored. Examples of computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage, and the like, and may also be implemented in the form of a carrier wave (for example, transmission over the Internet). Include. The computer readable recording medium can also be distributed over network coupled computer systems so that the computer readable code is stored and executed in a distributed fashion.

So far I looked at the center of the preferred embodiment for the present invention. Those skilled in the art will appreciate that the present invention can be implemented in a modified form without departing from the essential features of the present invention. Therefore, the disclosed embodiments should be considered in descriptive sense only and not for purposes of limitation. The scope of the present invention is shown in the claims rather than the foregoing description, and all differences within the scope will be construed as being included in the present invention.

As described above, since the linear relationship between the objects is given, the calculation is simple, and even when a change in the scene occurs, high accuracy can be maintained.

Claims (11)

(a) counting the number of zeros present in the DCT coefficients for the macro block; And (b) if the number of counted zeros is large, reducing the number of bits to be allocated in quantization, and if the number of counted zeros is small, increasing the number of bits to be allocated in quantization. Bit allocation method used in execution. The method of claim 1, wherein performing quantization is performed. A method of allocating bits used in performing quantization, characterized in that used in MPEG-2 or MPEG-4. (a) receiving a DCT coefficient for the macro block; (b) counting the number of zeros among DCT coefficients in the input macroblock and calculating a ratio ρ (i, j) value of the number of zeros; (c) determining whether the value of ρ (i, j) is calculated for all macroblocks of one image frame; And (d) if the macroblock to be input remains as a result of the determination and receives the next macroblock, and if all macroblocks are received, performing a predetermined bit allocation using the value of ρ (i, j). And a bit allocation method used for performing quantization. The method of claim 3, And the value of ρ (i, j) is calculated by the following equation. Here, N _zero (A) is the number of cases where the DCT coefficient value is 0 in one macroblock, and N (A) is the total number of DCT coefficients in the one image frame. The method of claim 3, wherein step (d) If the value of ρ (i, j) is large, the number of bits allocated is small. If the value of ρ (i, j) is small, the number of bits allocated is large. . 4. The method of claim 3, wherein performing quantization is A method of allocating bits used in performing quantization, characterized in that used in MPEG-2 or MPEG-4. An image input unit configured to sequentially input DCT coefficients of one macro block of one image frame in macroblock units; A counting unit for counting a number of zeros of DCT coefficients in the input macro block; An intermediate value calculator for calculating a ratio p (i, j) of the number of zeros counted by the counter; And And an allocating bit determining unit for linearizing using a value of ρ (i, j) calculated by the intermediate value calculating unit to determine bits to be used for quantization. The method of claim 7, wherein And the value of ρ (i, j) is calculated by the following equation. Here, N _zero (A) is the number of cases where the DCT coefficient value is 0 in one macroblock, and N (A) is the total number of DCT coefficients in the one image frame. The method of claim 7, wherein the allocation bit determining unit If the value of ρ (i, j) is large, the number of bits allocated is small. If the value of ρ (i, j) is small, the number of bits allocated is large. . 8. The method of claim 7, wherein performing quantization is An apparatus for allocating bits for use in performing quantization, characterized in that used in MPEG-2 or MPEG-4. A computer-readable recording medium having recorded thereon a program for executing the method according to any one of claims 1 to 6.