KR19990008989A - Apparatus and method for shape bit control in moving picture encoder - Google Patents

Abstract

The present invention relates to shape bit control in an image encoding method, and in particular, to control the degree of redundancy of image signals in a time domain to control the amount of shape bits generated during actual protection. In consideration of the generation amount and the target bit amount of the shape image, the target bit amount determination unit 110 for allocating the target bit amount of the current target image to be encoded for each object plane unit and each image block unit of a specific size; ; The threshold value for the prediction error size can be allocated to each object plane unit in consideration of the actual bit generation amount of the previous encoded image, the target bit amount of the target image to be encoded, and image quality deterioration characteristics. A predictive error threshold (TH) determiner 120 for controlling; A quantization interval size (QP) determiner that controls bit generation for each image block by defining a degree of compression for each image block in consideration of a target bit amount of the image block when quantizing a target image to be encoded in a block unit of a specific size. It relates to a shape bit control apparatus and method in a moving picture encoder configured to include (130).

Description

Apparatus and method for shape bit control in moving picture encoder

The present invention relates to shape bit control in a video encoding method, and more particularly, to a shape bit control apparatus in a moving picture encoder that controls the degree of redundancy of video signals in a time domain to control the amount of shape bits generated during actual coding. It is about a method.

In general, an image encoding technique aims to minimize the amount of encoded image data (Bitstream) while minimizing distortion of the image quality of a reconstructed image.

In order to achieve the above object, the image quality of the reconstructed image is obtained by using the correlation between the image data between the previous image and the current image in the time domain and the correlation between the image data in the spatial domain. A compression encoding method that reduces the amount of video data to be encoded while minimizing the degree of degradation is used.

In addition, one of the important techniques to be considered in such an image encoding method is a technique capable of controlling the amount of coded shape bits.

This technique allows the encoder to generate the amount of video information to be compressed and encoded according to a transmission path having a limited bandwidth, so that a time delay for reconstruction is not generated when the decoder restores the image.

Conventionally, in order to encode according to a target bit amount allocated to an image to be encoded, a bit control method of controlling an encoding bit amount by encoding an image by considering only a quantization interval is mainly used.

In the conventional digital image encoding method, as shown in FIG. 1, a method of improving encoding efficiency by using correlation with input image data generally uses two encoding techniques.

The first method converts input image data to be coded from a temporal frequency domain to a spatial frequency domain using a discrete cosine transform (DCT). Transform Coding is used to define the correlation between data.

This encoding method is characterized in that it extracts only the spatial frequency characteristics of the image data without losing the image data, and uses it for the compression and encoding method of the image, and is widely used in still image encoding methods.

The second method is a predictive coding method using differential pulse code modulation (DPCM), which defines correlation of image data to be encoded in a time domain.

As shown in FIG. 1, when the image input device 1 receives the nth coded image O, the motion degree of each encoded image unit is moved compared to the n-1th coded previous image. The estimation apparatus 2 estimates using the vector value, and the motion compensation apparatus 3 configures the prediction image P for the n-th original image by the estimated motion vector.

The example image configured as described above has an error due to movement when compared with the actual nth original image. The error due to such movement is called a prediction error (e) and is defined as Equation (1).

[Equation 1]

e = O-P

The prediction error defined as described above means the degree of correlation of the image data in the time domain. Since only the image data configured by the prediction error is used by the real encoder 4, there is a loss of the image data, but the coding efficiency of the image is reduced. It has the characteristics to be improved.

Therefore, in the moving picture coding method, as shown in FIG. 1, a hybrid coding method using both a discrete cosine transform (DCT) extracting spatial correlation of an image signal and a differential pulse code modulation (DPCM) method extracting temporal correlation ( Hybrid Coding) is used a lot.

However, since the image information to be encoded as described above is configured to determine the actual bit generation amount only by the quantization interval size of the quantizer, there is a problem that the correlation between the bit generation amount and the reconstructed image quality is relatively high.

Accordingly, the present invention was devised to solve the above-mentioned problems. In the moving picture encoder, it is possible to control the amount of shape bits in actual encoding by controlling the degree of redundancy of video signals in the time domain. It is an object of the present invention to provide a shape bit control apparatus and method.

1 is a system configuration diagram of a general video encoder;

2 is an overall system configuration diagram of a video encoder for a shape bit control apparatus according to the present invention;

3 is a block diagram of a shape bit control apparatus according to the present invention;

4 is a flowchart illustrating an operating state of the prediction error threshold determination unit of FIG. 3;

5 is a flowchart illustrating an operating state of the quantization interval size determiner of FIG. 3.

Explanation of symbols for the main parts of the drawings

100: shape bit control device

110: target bit amount determination unit

120: prediction error threshold determination unit

130: quantization interval size determination unit

111: target bit amount determination unit for each image object plane

112: target bit amount determination unit for each image block

According to the present invention for achieving the above object, in consideration of the actual bit generation amount of the shape image and the target bit amount of the shape image, the target bit amount of the target image to be currently encoded is determined by an object plane. A target bit amount determiner 110 for allocating each unit and each image block unit having a specific size; The amount of encoding shape bits can be allocated by assigning a threshold value for the prediction error size to each object plane unit in consideration of the actual bit generation amount of the previous encoded image, the target bit amount of the target image to be encoded, and image quality deterioration characteristics. A prediction error threshold (TH) determiner 120 for controlling the; When quantizing a target image to be encoded in a block unit of a specific size, the quantization interval size (QP) that controls the generation amount of bits per image block is determined by defining the degree of compression for each block in consideration of the target bit amount of the image block. Characterized by including the unit 130.

In addition, the present invention for achieving the above object, the target bit amount determination process (S10) for allocating the target bit amount of the image object plane to be encoded, for each image block unit of the image object plane and a specific size; A prediction error threshold determination process (S20) of constructing a prediction image P by allocating a prediction error threshold according to a target bit amount; A quantization interval size determination process (S30) is performed by assigning a quantization interval size (QP) to each image block according to a target bit amount of an image block having a specific size and assigning the intra image block in the prediction error image (e). It is characterized by including the.

As shown in FIG. 2, in order to encode an image to be encoded according to an allocated target bit amount, the quantizer 4 adjusts the quantization interval size QP in units of image blocks, and then, A bit control method for controlling the amount of shape encoding bits by compression and encoding, and a predictive encoding process having an effect of reducing the amount of image information to be encoded by extracting image signals overlapping between images in a time domain using motion information between images. By simultaneously performing a method of controlling the shape bit generation amount by reconstructing the predicted video block P constructed using the motion information of the motion compensation device 3 using the threshold value TH for the prediction error e, The amount of shape coded bit generation can be controlled more efficiently.

In other words, when the shape bit is controlled in the image object plane unit (the image unit having characteristics such as motion, shape, and structure), the quantizer 4 is used to compress the actual image. In addition to the bit control method based on the quantization interval used, a method of reducing the amount of video information to be encoded by extracting the redundancy of the video signals in the time domain using motion information between the images is performed. The degree of redundancy is controlled to control the amount of shape bits generated during actual encoding.

The target bit amount determination unit 110 determines the target bit amount of the target image to be encoded currently and the unit of the image object plane in consideration of the actual bit generation amount of the shape image and the target bit amount of the shape image. Each image block unit of size is allocated.

In the prediction encoding method, in order to reduce the amount of information of the target image to be encoded, the difference image P information processed by the actual encoder is differently defined according to the magnitude of the prediction error (e) generated when the prediction method using the motion information is used. Can be.

The prediction error threshold determiner 120 determines the actual bit generation amount of the previous image in which the threshold value for the prediction error magnitude is encoded, the target bit amount and the image quality degradation characteristic of the target image to be encoded, according to the above characteristics. In consideration of this, the generation of the coded shape bit is controlled by configuring the allocation by the unit of the image object plane.

When the quantization interval size (QP) determiner 130 quantizes a target video to be encoded in units of blocks having a specific size, the compression level of each video block is defined in consideration of the target bit amount of the video block, thereby making it possible to determine the video block wall. Control bit generation amount.

Herein, the operation principle of the shape bit control apparatus in the video encoder according to the object of the present invention will be described in detail.

First, the operation principle of the target bit amount determination unit 110 will be described.

In the image encoding method, the image quality and the bit generation amount of an encoded image are determined according to the degree of quantization.

The method of determining the degree of quantization is considered to be a very important factor, which is basically defined in inverse proportion to the target bit amount of an image to be encoded.

Therefore, the target bit amount allocation is a condition that must be preceded in the quantization process.

In FIG. 3, the target bit amount determiner 110 is a processor that allocates a target bit amount of a target video to be encoded for each image object plane unit and image block unit, and includes a total bit rate and an actual bit generation amount ( Buffer).

A target bit amount of each image object plane (hereinafter referred to as PTB) is defined as in Equation 2.

Where N is the total number of images per second to be encoded, Rn is the number of residual images to be encoded, PAB (Actual Bits of Plane) is the actual amount of bits generated per image object plane, SPAB (Sum of PAB) is the amount of accumulated bits, RPTB (Remain of PTB) represents the remaining target bit amount.

[Equation 2]

RPTB = BitRate-SPAB

PTB = RPTB / Rn, (0≤Rn≤N)

However, when RnN and RPTB ≤ 0 are satisfied, since there are no more bits to allocate to the target video to be encoded, the target video to be encoded is not encoded until the PTB0 is satisfied.

A target bit amount (hereinafter referred to as BTB) for each video block is defined as in Equation (3).

That is, it is determined according to the number of intra-coded blocks (referred to as coding blocks hereinafter CB) except intra (intra: motion blocks cannot be encoded using only motion information and weak all blocks except 0 or 255). do.

Equation 4 is a relation for predicting the number of intra image blocks.

[Equation 3]

BTB = (PTB-SBAB) / (CB-nB), (0≤nB≤CB)

[Equation 4]

CB = (AB * PB (TH) / (PPB, (PPB0, PB (TH) 0)

Ratio = PTB _t / PAB _t-1 , (0≤t≤N)

Here, SBAB (Sum of BAB) is the cumulative bit generation amount,

nB is the number of encoded video blocks,

TH is the prediction error threshold,

AB is the number of actual intra image blocks participating in the quantization process in the previous image object plane,

PB (TH) is a number of intra-image blocks predicted by considering only TH values without considering motion information in a target image object plane to be encoded,

PPB represents the PB (TH) value used in the previous image object plane.

Ratio is a value that determines the relationship between the actual bit generation amount of the previous image object plane and the target bit amount of the image object plane to be encoded. The ratio of the predictive intra image blocks is determined by using an image characteristic that the image data are similar in the time zone. Used to predict more accurately.

Second, the operation principle of the prediction error threshold determiner 120 is as follows.

In the video encoding method, a prediction encoding method is used to reduce the amount of bits generated.

That is, as illustrated in FIG. 2, the prediction image P defined by using motion information between the original image O and the previous reconstructed image R is determined, and the determined prediction image P and the original image O are determined. A prediction error image (e) representing a difference between and is obtained, and a method of actual quantization thereof is used.

A factor considered in such a predictive method is a prediction error threshold TH that is used as a decision criterion when the prediction image P is correct.

That is, when determining whether or not the original image O can be defined using only the previous reconstructed image R, a total error Perr of the prediction error image e defined using motion information and a prediction error threshold ( TH) value is compared with each other and judged.

Equation 5 is a conditional expression indicating the above relationship.

[Equation 5]

if (Perr ≥ TH) P = 0

elseP = R

In the above relation, the prediction error threshold TH has a function of adjusting the bit generation amount by adjusting the number of intra video blocks.

Accordingly, the prediction error threshold determination unit 120 may adjust the encoding bit amount and the image quality of the shape image by using the prediction error moon touch characteristic as shown in Equation 5 above.

The prediction error threshold TH has a function of adjusting the number of intra blocks in image encoding.

Therefore, when the bit generation amount is increased, the prediction error threshold value TH should be reduced, and in the opposite case, the prediction error threshold value TH should be increased.

As described above, in order to decrease / increase the value of the prediction error threshold TH, it is necessary to know in advance the correlation between the bit rate of change and the prediction error threshold TH.

Table 1 defines this relationship. For example, if TH (= PrevTH) of the previous image object plane is 48 and you want to generate a bit that is 25% higher than the current bit generation amount, It can be adjusted to 80 ”.

TABLE 1

TH Adjustment Look-Up Table

In FIG. 4, TR has a value equal to Ratio defined in Equation 4, PrevTH is a TH value used in the previous image object plane, and PrevQP is a quantization parameter used in the previous image object plane (hereinafter referred to as QP). Defined as the average of the values.

In addition, FIG. 4 illustrates a case in which a quantization interval size (QP) value is limited to only three values and the threshold value (TH) is configured as 8 bits when quantizing a shape image. Is showing.

TR is an information on how much to reduce / increase the bit generation amount of the current object plane to be encoded by comparing with the bit generation amount of the previously encoded image object plane. Adjust the TH value using Table 1 below.

Meanwhile, when adjusting TH in FIG. 4, one more factor of PrevQP is considered, which was used for the purpose of preventing image quality deterioration.

Here, the QP value is a quantization interval size value that defines how much to quantize the image block, PrevQP is an average value of the QP values for each image block used in the previous image object plane, when having three values of 1, 2, 4 Only shown in FIG. 4, and if the QP has a variety of values, it can be extended by applying a method similar to FIG.

In addition, the look-up table in FIG. 4 uses two parameters, the first of which is defined as PrevTH, and the second of which is an allowable bit change rate.

For example, when PrevTH = 48 and PrevQP = 1 and TR = 0.4 in FIG. 4, the maximum allowable bit change rate is defined as 0.25, which means a bit change rate obtained only by TH adjustment, and the remaining 0.15 bits The change is configured to process the QP value defined in the quantization interval size determiner 130, so that the system is configured to satisfy the final bit generation rate of 0.4.

Thus, the adjusted TH is defined as 128.

As described above, since TH is configured to vary the degree of adjustment according to the previous QP value, the threshold value TR in FIG. 4 may be classified according to the type of quantization interval size QP used. have.

Third, when the quantization interval size (QP) determiner 130 quantizes a target video to be coded in units of blocks of a specific size, the quantization interval size is taken into consideration of the degree of compression for each video block in consideration of the target bit amount of the video block. Defined by the size (QP) value, the processor for controlling the bit amount per shape image block determines the degree of quantization of the intra image block.

In the present invention, a method of defining a quantization interval size when using only three quantization interval size (QP) values of quantization 1, 2, and 4 is discussed as an example.

This method can be extended even when the quantization interval size (QP) has various values.

Equation 6 below shows a change amount of a buffer using a relationship between a target bit amount (PTB) per image object plane and the number of intra image blocks (CB) to be encoded, and FIG. 5 shows a quantization interval size (QP) using this value. Shows the operation of the quantization interval size determiner 130 to determine the value.

[Equation 6]

DR = ((SBAB / PTB)-(nB / CB) * CB

BR = DR * CB

BR defined in Equation 6 is a value that defines the excess / decrease rate for the target bit amount (BTB) of an image block defined in Equation 3, and when BR is greater than zero, the quantization interval size (QP) is increased. Value, and vice versa.

The QP value is adjusted in the intra image block unit and determines the actual quantization degree.

Participating in the adjustment of this value, the element is the actual bit generation amount and the number of intra blocks, and the QP value is defined according to the BR value representing the ratio between them.

Also, the adjusted QP value is adjusted based on the average QP value (= TQP) used in the previous image object plane. When BR is negative as shown in FIG. 5, that is, the actual bit generation amount is smaller than the target bit amount. When it occurs, the QP is reduced as shown in FIG. 5 according to the size of the value, so that the bit generation amount of the current video block to be encoded is increased, and when the BR has a positive value, the QP value is increased in the same manner, The bit generation amount of the video block to be encoded is reduced.

On the other hand, the reason for using QP is to provide a QP change criterion so that a sudden change (for example, 1-4, 4-1) between adjacent QPs does not occur, and the image information of images in the adjacent domain is The purpose of the present invention is to provide a quantization method for reducing the degree of deterioration in image quality of a reconstructed image by defining the degree of quantization similar to the previous image according to similarity.

As described in detail above, the present invention compresses and encodes video signals by adjusting a quantization interval size (QP) in units of image blocks in order to encode shape image information to be encoded according to an allocated target bit amount. A bit control method for controlling the amount of shape coding bits, which extracts video signals overlapping between images in a time domain using motion information between images, and uses motion information in a prediction encoding process having an effect of reducing the amount of image information to be encoded. By parallelizing the shape bit control method of controlling the prediction error threshold value TH, which defines the number of image blocks participating in the actual quantization process among the prediction image blocks P, in units of an image object plane, In addition to controlling the amount of shape encoding bit generation more efficiently, the quality of the image It has the advantage of being able to provide the effect to lower than compared to the degree of the shape encoding method using only the QP.

In addition, it has the advantage that it can be effectively applied to the low bit rate coding method.

Claims (6)

In consideration of the actual bit generation amount of the shape image and the target bit amount of the shape image, the target bit amount of the current image to be encoded is allocated to each image object unit and image block unit of a specific size. A target bit amount determining unit 110; Receives the previous reconstructed image R from the original image O input from the image input apparatus 1 and the image reconstruction apparatus 6, and the motion information and the prediction error threshold TH extracted from the motion estimation apparatus 2 When the prediction image P is defined in the motion compensation apparatus 3, the threshold value of the prediction error size is determined in consideration of the actual bit generation amount of the previous encoded image, the target bit amount of the target image to be encoded, and image quality deterioration characteristics. A prediction error threshold (TH) determiner 120 for controlling the amount of coded bit generation so as to be allocated in units of an image plane; When the prediction error threshold determiner 120 quantizes the intra target image to be encoded in a block unit having a specific size, in the prediction image P configured according to the prediction error threshold TH determined based on the shape coding bit amount Determining the Quantization Interval Size (QP) to Control the Generation of Bits per Image Block by Allocating the Quantization Interval Size (QP) Indicative of the Quantization Level for Each Image Block by Each Shape Image Block Unit Considering the Target Bit Size of the Image Block And a unit (130). A shape bit control apparatus in a moving picture encoder. The method of claim 1, The target bit amount determination unit 110, A target bit amount determiner 111 for each image object plane that determines the target bit amount of the target image object plane to be encoded using the total target bit amount and the actual bit generation amount SPAB; The number of intra image blocks to be quantized is defined as the number of intra image blocks (AB) actually occurring in the previous image object plane to be encoded and the number of intra blocks (PB) defined using only the prediction error threshold value (TH) without motion information in the image object plane to be encoded. And a target bit amount determiner 112 for each video block to predict the target bit amount of the video block to be quantized using the ratio of the ratio and the target bit amount to the actual bit generation amount. And a shape bit control device in a moving picture encoder. A target bit amount determination process (S10) of allocating a target bit amount of an image object plane to be encoded for each image block unit and a specific size of image block unit; A prediction error threshold determination process (S20) of constructing a prediction image P by allocating a prediction error threshold according to a target bit amount; A quantization interval size determination process (S30) of quantizing an intra image block in a prediction error image (e) by allocating a quantization interval size (QP) to each image block according to a target bit amount of an image block having a specific size. And a shape bit control method in a moving picture encoder. The method of claim 3, The target bit amount determination process (S10), RPTB = BitRate-SPAB PTB = RPTB / Rn, (0≤Rn≤N) BTB = (PTB-SBAB) / (CB-nB), (0≤nB≤CB) CB = (AB * PB (TH) / PPB, (PPB0, PB (TH) 0) Ratio = PTB _t / PAB _t-1 , (0≤t≤N) The shape bit control method of a moving picture encoder, characterized in that using the same method. The method of claim 3, The prediction error threshold determination process (S20), if (Perr ≥ TH) P = 0 elseP = R The shape bit control method of a moving picture encoder, characterized in that using the same method. The method of claim 3, The quantization interval size determination process (S30), DR = ((SBAB / PTB)-(nB / CB) * CB BR = DR * CB The shape bit control method of a moving picture encoder, characterized in that using the same method.