KR100273459B1 - Bit string control method of image data - Google Patents

Abstract

The present invention divides the image data into several blocks on a frame and efficiently assigns a quantization step value in a system for differentially encoding by varying weights according to importance, so that each quantization step size value for each block separated on the frame is different. Obtaining a first step; Based on the obtained quantization step size value, the bit amount to be generated for each block is obtained, and the quantization step size value is adjusted to the maximum change width allowed for the boundary block or the object block and the minimum change width allowed for the background block. Step 2; A third step of obtaining the quantization step size of the next block using the amount of bits generated when encoding based on the obtained quantization step size value; The generated bit amount of the current frame used for encoding the next frame is processed through the fourth step of replacing the actual bit amount with the bit amount actually used to obtain the quantization step size of the next frame.

Description

Bit string control method of image data

The present invention relates to a technique of dividing the image data into several blocks on a frame-by-frame basis and encoding them differently according to their importance. Particularly, when weighting each divided block, the quantization step size is efficiently obtained. The present invention relates to a method of controlling bit streams of image data in which a boundary block and an object block are divided to adjust a magnitude value of a quantization step and then a quantization step size value of a next block.

Currently, most video encoding / decoding techniques such as H.263 and MPEG-4 use a method of adjusting a quantization step value to control a bit string.

For example, in the case of TMN-10, a test model of H.263 version-2, the difference between the generated bit amount of the previous frame and the target bit amount of the previous frame, the target bit amount up to the previous block of the current frame, The amount of bits generated up to the actual previous block is calculated to determine the value of the quantization parameter (QP) of the current frame and block, and the actual amount of bits is continuously adjusted.

Recently, various methods for controlling the allocation rate of coded bits have been studied, but basically satisfy various constraints such as target bit rate, delay time, image statistics, etc. The amount of bits generated by the method used for encoding is determined by determining a quantization parameter value having a minimum distortion.

In most cases, in order to improve coding efficiency at a low bit rate or to limit a sudden change in inter-block quality, a change width of the quantization parameter between adjacent blocks is limited.

In addition to rate control methods, video coding techniques such as video conferencing and video communication have been conducted to separate frames into important blocks such as people and less important blocks, and to encode them in different ways.

FIG. 1 illustrates an example in which encoded data is divided into a background block (1), a boundary block (2), and an object block (3) in units of frames in a conventional encoding system. FIG. An example of weighting according to a block is shown. FIG. 3A is a bit amount allocation table when no weight is assigned according to a block, and FIG. 3B is a bit amount allocation example table when a weight is assigned according to a block according to the prior art.

Hereinafter, an example of a bit string control method according to the prior art will be described with reference to FIGS. 1 to 3.

First, one image frame is composed of 10 blocks, which are composed of six background blocks (1) and four important blocks, that is, two boundary blocks (2) and two object blocks (2). .

Second, the quantization step of the first block was determined to be 10 using the encoding result of the previous frame.

Third, the variation range of the quantization step value is limited to +2 and -2.

Fourth, the target bit amount of the current frame is 100 bits, 7 bits are allocated to the background block 1 in consideration of the weight, and 13 bits are allocated to the critical blocks 2 and 3.

Fifth, a linear bit amount occurs according to the quantization step value. That is, each time the quantization step value increases (decreases) by 1 bit, the amount of generated bits decreases (increases) by one.

Sixth, the quantization step value is adjusted by changing the range of +2 and -2 by using the difference between the target bit amount and the generated bit amount.

The result of controlling the bit string in this manner is shown in FIG. 3B. As shown in this result table, limiting the variation of the quantization step value between adjacent blocks simply adds weights to the blocks and bit string control methods, so that the actual boundary block (2) or the object block (3) can be used more. It can be seen that it is impossible to allocate a bit amount.

In particular, this phenomenon is more severe when the change in the quantization step size between adjacent blocks is smaller, and the image size is smaller or the block is wider. For example, in the bit stream of FIG. 3B, the quantization magnitude values of the critical blocks 2 and 3 are 16, 14, 12, and 10, respectively, but the actual generated bits are weighted 4, 6, 8, and 10, respectively. It can be seen that more bits have not actually been allocated as compared to not assigning bits.

As described above, when the bit string control method of image data according to the prior art is used, it is impossible to allocate more bits to important blocks, and the degree of change of the quantization step size between adjacent blocks is smaller. In other words, the wider the block, the more severe the result is a defect that can not provide a good image.

Therefore, the technical problem to be achieved in the present invention is to obtain the quantization step size value for each block, and then to obtain each parameter using the quantization step size value with the maximum change width for the boundary block or object block, the minimum change width for the background block In the present invention, a bit string control method is provided to obtain a quantization step size value of a next block by using a parameter obtained when encoding using the obtained quantization step size value.

1 is an explanatory diagram showing an example of region division of an image frame;

Figure 2 is an exemplary diagram given a weight for each block in the bit string control method according to the prior art.

3A is a coding process table when no weight is applied according to an area.

3B is a coding process table in the case where each weight is assigned by the prior art.

4 is a signal flowchart of a method for controlling bit strings of image data according to the present invention;

5 is a coding process table in the case where a weight is applied according to an area according to the present invention.

6 is an exemplary diagram for assigning weights for each block in the bit string control method according to the present invention;

*** Description of the symbols for the main parts of the drawings ***

1: Background block 2: Border block

3: object block

According to an aspect of the present invention, there is provided a method of controlling bit streams of image data, the method comprising: a first step of obtaining respective quantization step size values for each block separated on a frame; Based on the obtained quantization step size value, the bit amount to be generated for each block is obtained, and the quantization step size value is adjusted to the maximum change width allowed for the boundary block or the object block and the minimum change width allowed for the background block. Step 2; A third step of obtaining the quantization step size of the next block using the amount of bits generated when encoding based on the obtained quantization step size value; The generated bit amount of the current frame used for encoding the next frame is composed of a fourth step of replacing the actual bit amount with the actual generated bit amount to obtain the quantization step size of the next frame. 1 and 4 to 6, a process for encoding one frame will be described in detail with an example.

First, the quantization step size value of each block (1), (2), (3) is found by using various existing methods used to adjust the bit string.

For example, the first block of the current frame based on the amount of bits generated when encoding the discrete frame, that is, the total bit amount B _t _ _mod when the current block of the current frame is encoded using QP _i _ _mod . Find the quantization step size for. (S1)

This operation is generally expressed as calculating the quantization parameter (QP). The relationship between the quantization step size and QP is as follows.

Quantization Step Size = 2 × QP

In addition, various parameters and algorithms may be used to calculate the QP, but it is assumed here that the bit amount generated in the previous frame is used as one of them.

Using the quantized step size values thus obtained, the result of encoding is obtained. That is, the resultant bit amount _Bi is calculated when encoding using the quantization parameter QP _i of the i-th block (S6).

After that, the quantization step size value is adjusted to the maximum (+ max) change width for the boundary block 2 or the object block 3 and the minimum (-min) change width for the background block, and the result is actually encoded.

That is, the value of the changed quantization parameter QP _i _ _mod is obtained by using the previously obtained quantization parameter QP _i and the block information (important block, object block, boundary block) of the current block. (S2-S4)

Here, QP _i _ _mod refers to a quantization parameter modified in consideration of the region of the i-th block.

QP _i _ _mod = QP _i -D _{i where} object block (3) and boundary block (2)

= QP _i -D _i However, background block (1)

Here, the frame to the object block 3 and the boundary block (2), the background block (1) but separated in two steps, comprising two or more separate when there to adjust the weight degree of each block by appropriately adjusting the value of D _i of have. In general, in the case of the second stage, D _i = max (the maximum change in QP). Here, D _i means a change width (-max change width ≤ D _i ≤ + max change width) of the quantization parameter considering the region.

The quantization step size of the next block is obtained using the obtained parameter when encoding using the already obtained quantization step size value. That is, the amount of bits B _i _ _mod generated when the i-th block is encoded using the QP _i _ _mod is calculated (S5).

The generated bit amount of the current frame used for encoding the next frame is used to obtain the quantization step size of the next frame by replacing the generated bit amount with the actual generated bit amount.

That is, QP _{i + 1} is obtained for encoding the i + 1 th block, wherein the generated bit amount up to the present is not the bit amount B _{t_mod} generated during actual encoding, but the current block of the current frame. Until now, the total bit amount B _{t in} the case of encoding using QP _i is used (S9). Therefore, the value of the _{i +} 1th QP _{i + 1} can be calculated independently of the weight coding using any bit string adjustment method using the amount of generated bits as a parameter.

When the encoding is performed in the ninth step S9, a difference between the actual generated bit amount and the calculated bit amount after encoding one frame is obtained. Obtaining the quantization step size of the first block in the next frame The difference is compensated for by using the amount of bits actually generated.

An example of the result of encoding the input data through the above process is shown in FIG. 5, and FIG. 6 exemplarily shows the quantization step size value of each block.

As described in detail above, the present invention obtains the quantization step size value for each block, and then obtains each parameter using the quantization step size as the maximum change width in the case of the boundary block or the object block and the minimum change width in the case of the background block. By adjusting the value and encoding the obtained quantization step size value, the quantization step size of the next block is obtained by using the obtained parameter. The optimal quality is applied to a video system such as video conferencing. There is an effect that can be done.

Claims (2)

Obtaining a quantization step size value for each block separated on the frame; Based on the obtained quantization step size value, the bit amount to be generated for each block is obtained, and the quantization step size value is adjusted to the maximum change width allowed for the boundary block or the object block and the minimum change width allowed for the background block. Step 2; And a third step of obtaining the size of the quantization step of the next block by using the amount of generated bits obtained when encoding based on the obtained quantization step size value. The image of claim 1, further comprising: replacing the generated bit amount of the current frame used for encoding the next frame with the bit amount actually generated to use to obtain the quantization step size of the next frame. Bit string control method of data.