KR100214584B1 - Method and apparatus for alloting bit quantity of variable rate moving image encoder - Google Patents

Abstract

The present invention relates to a technique for allocating a bit rate based on image quality in a variable rate video encoding technique. In the conventional variable rate encoding scheme, a bit to be used in an actual variable rate encoder based on a bit amount generated after using a fixed quantizer. Since the amount is predicted and allocated, there is a defect that cannot provide a uniform picture quality. In order to solve the problem, the present invention provides a first step of encoding a fixed amount of bits generated per unit time with respect to an input image at a predetermined value. And a second step of evaluating and storing the image quality of the reconstructed frame with respect to the input frame generated during the fixed rate encoding, and confirming that the fixed rate encoding of the entire image is terminated, based on the stored image quality evaluation value. A third step of calculating and storing a target bit amount for each unit time; and stored in an output buffer Through the fourth step, the bit rate error is calculated based on the target bit amount by time, and the bit rate controller adjusts the size of the quantization step based on the bit rate error and the complexity in the image so that the bit amount generated per unit time matches the target value. Bit amount is allocated.

Description

Method and apparatus for allocating bit rate in variable rate video encoder

1 is a block diagram of a typical variable rate video encoder.

2 is a block diagram of a bit rate allocation apparatus of a variable rate video encoder of the present invention.

3 is a graph showing a bit rate allocation method according to the present invention.

4 is a signal flow diagram of the bit amount allocation method of the present invention.

* Explanation of symbols for main parts of the drawings

101: subtractor 102: DC

103: quantizer 104: local decoder

105: adder 106: interframe predictor

107: variable length encoder 108: buffer

109: activity detector 110: bit rate regulator

121: image quality detector 122: hierarchical bit rate allocator

The present invention relates to a technique of allocating a bit rate based on image quality in a variable rate video encoding technique. In particular, the present invention relates to predicting and assigning a bit amount to be used for actual variable rate encoding based on an image quality evaluation of a resultant image of a fixed rate encoding process. A bit rate allocation method and apparatus for a variable rate video encoder.

A video encoder is roughly classified into a constant bit rate (CBR) coding method and a variable bit rate (VBR) coding method according to characteristics of an output bitstream. This depends on whether the amount of bits generated per unit time (eg 1 second) is fixed or variable.

The fixed rate coding method is preferred in an environment in which data is transmitted by using a channel having a fixed capacity, such as TV broadcasting, but the same bit amount is allocated to images having different complexity. The disadvantage is that the image quality is uneven.

On the other hand, the variable rate coding scheme allocates bits in proportion to the complexity of the image, which has the advantage of maintaining a constant picture quality over the entire playback period. : Digital Video Disc) has been put to practical use in application fields.

1 is a block diagram of a typical two-step variable rate coding scheme. As shown in FIG. 1, the main image encoder 100 generally used in the fixed rate coding scheme and a target bit rate varying according to the complexity of an image are illustrated in FIG. The bit rate allocator 111 for supplying the target bit rate is configured in a combined form, and its operation will be described as follows.

The subtractor 101 generates a prediction error C, which is a difference between the input frame A and the prediction frame B, and the prediction error C is a discrete cosine transform in the DC unit 102. And then quantized in quantizer 103.

The local decoder 104 and the adder 105 add the quantized signal E and the prediction signal B to generate a decoded signal G, and the interframe predictor 106 generates the decoded signal G. ) Generates a prediction frame B corresponding to the frame A to be input next.

In addition, the quantized signal E is converted into a variable length code through the variable length coder (VLC) 107 and then transmitted through the buffer 108.

The bit rate adjuster 110 is configured between the complexity (I: activity) of each macro block (MB: Macro-Block) in the frame obtained by the activity detector 109 and the target bit amount calculated in the buffer 108 and the generated bit amount. The amount of bits generated per unit time is adjusted by adjusting the size of the quantization step based on the difference (J).

In the fixed rate coding scheme, the input video signal VIDEO _in is encoded using only the above blocks, wherein the target bit amount L per unit time based on the buffer 108 is always constant.

The bit rate allocator 111 calculates a variable target bit amount for each unit time, and the main image encoder 100 performs variable rate encoding using the same.

The variable rate encoding process will be described step by step as follows.

First step: After fixing the size K of the quantization step to a specific value, encoding is performed to calculate and store the amount of bits generated per unit time. After encoding the entire image, the target bit amount L for each unit time is calculated and stored based on the stored bit amount. In this case, the target bit amount has a variable value depending on the complexity of the image.

Second step: The target bit amount L for each unit of time is supplied to the buffer 108, and the buffer 108 generates a bit rate error J based on this, and the bit rate adjuster 110 generates the bit rate error J and the complexity I in the image. By adjusting the size of the quantization step with reference to the amount of bits generated per unit time to match the target bit amount. That is, variable rate encoding is performed. However, in the conventional variable rate coding scheme, since the variable rate encoder is adjusted based on the amount of bits generated after using a fixed quantizer, that is, a method of indirectly reflecting image quality elements, it is impossible to provide uniform image quality. There was a fault. Accordingly, an object of the present invention is to provide a bit amount prediction allocation method and apparatus for obtaining a high quality characteristic by predicting a bit amount to be used for actual variable rate encoding based on an image quality evaluation of a result of a fixed rate encoding process.

According to an aspect of the present invention, there is provided a method for allocating a bit amount of a variable rate video encoder, the method comprising: encoding a fixed bit amount generated per unit time with respect to an input image at a predetermined value (S1); A second step (S2) of evaluating and storing the image quality of the reconstructed frame with respect to the input frame generated during the fixed rate encoding; A third step (S3, S4) of checking that the fixed rate encoding for the entire image is finished and calculating and storing a target bit amount for each unit time based on the stored image quality evaluation value; The bit rate error is calculated based on the target bit amount per unit time stored in the output buffer, and the bit rate controller adjusts the size of the quantization step based on the bit rate error and the complexity in the image to make the bit amount generated per unit time match the target value. It consists of four steps (S5-S7).

2 is a block diagram of a bit amount prediction allocation apparatus of the present invention, as shown therein, a subtractor 101 for generating a prediction error C which is a difference between an input frame A and a prediction frame B, The DC unit 102 for discrete cosine transforming the prediction error C, the quantizer 103 for quantizing the discrete cosine transformed image, and the quantized signal E and the prediction signal B are added up. A local decoder 104 and an adder 105 for generating a decoded signal G, and an interframe predictor for generating a prediction frame B 'corresponding to a frame A' to be input next from the decoded signal G. 106, a variable length encoder 107 which receives the quantized signal E, converts it into a variable length code, and supplies it to the buffer 108 at the output stage, and the complexity of each macroblock in the frame obtained by the activity detector 109. (I) and the target bit amount calculated in the buffer 108 and the generated bit amount In the main image encoder 110 configured of the bit rate adjuster 110 for adjusting the size of the quantization step based on the difference J, the reconstruction is performed based on the input frame A generated while the fixed rate encoding is performed. Hierarchical bit rate hierarchically assigning a target bit amount proportional to the complexity of the image according to the image quality detector 121 for evaluating the image quality of the frame and the image quality evaluation value of the reconstructed frame output from the image quality detector 121; It will be described in detail with reference to Figures 3 and 4 attached to the operation and effect of the present invention configured as described, including the allocator 122 as follows.

The purpose of the variable rate coding technique is to allocate bit amounts according to the complexity of an image so that the image quality is maintained uniformly over the entire reproduction period. That is, the average bit rate calculated for a relatively long time (eg 1 hour) is kept constant, so that the total bit rate is kept constant, and the bit rate generated per unit time (eg 1 second) is variable according to the complexity of the image. This method is to realize high quality characteristics by assigning and encoding.

The present invention proposes a method for calculating a target bit amount by using the reproduction quality itself of each video as its complexity. Each block belonging to the main image encoder 100 in FIG. 2 plays the same role as that in FIG. 1, and in FIG. 2, a quality detector 121 and a hierarchical bit rate allocator 122 are shown. ) Is a block for calculating a target bit amount proportional to the complexity of an image, like the bit rate allocator 111 in FIG. 1, to replace the bit rate allocator 111. The method of calculating the target bit rate for each unit time used in the present invention is as follows.

First, in a first step, a fixed rate encoding is performed in which a bit amount generated per unit time is fixed to R _avg for an input image. Therefore, the target bit amount L input into the buffer 108 always has a constant value.

Second, the image quality M of the reconstructed frame G is evaluated and stored for the input frame A generated while the fixed rate encoding is performed, and then supplied to the hierarchical bit rate allocator 122. In this case, as a measure of image quality evaluation, a mean of absolute difference (MAD), a mean of squared difference (MSD), a signal to noise ratio (SNR), etc., may be used.

Third, when the one-step fixed rate encoding of the entire image is completed, the target bit amount L for each unit time is calculated based on the stored image quality and supplied to the buffer 108 during variable rate encoding. At this time, the target bit amount is allocated by the hierarchical bit rate allocation method as follows.

The bit amount R generated during encoding and the image quality D (assuming that MSE is used as the image quality scale for convenience of description) have a non-linear relationship as follows by the rate-distortion theorem.

Here, the bit evaluation function B (D) is defined as follows.

Referring to the first bit rate allocation, as shown in the hierarchical bit rate allocation graph shown in FIG. Calculate the evaluation function B _ti . Here, B _ti is defined as the sum of the bit evaluation functions for each frame included in the section _ti . The bit amount C _ti for each section is primarily allocated as follows based on B _ti .

Here, B _T is a bit amount for the entire predetermined image.

Referring to the second bit rate allocation, as shown in FIG. 3, the bit evaluation function for each time unit is performed after subdividing each section into a relatively short time unit (eg, 1 second) based on the first allocated bit amount Cti. Calculate Bsi '. Here, Bsi 'is defined as the sum of the bit evaluation functions for each frame included in the section si, and the bit amount Rsi of each section si is obtained as follows.

Here, f (·) is a function for calculating the bit amount of each section from the bit evaluation function can be used as the following equation.

In this equation, max (x, y) is a function that selects the larger value of x, y, and Rmas means the maximum bit rate allowed. After substituting Equation 6 into Equation 4, the optimum coefficients K and L satisfying the condition of Equation 5 are obtained. Then, after finally substituting these coefficients into (Equation 4), the bit amount Rsi of each section si is obtained and stored.

Fourth, the stored target bit amount L per unit time is supplied to the buffer 108 so that the buffer 108 generates a bit rate error J based on this, and the bit rate adjuster 110 generates the bit rate error J and the complexity I in the image. By adjusting the size of the quantization step as a reference, the amount of bits generated per unit time coincides with the target amount of bits. That is, the encoding of the variable is performed.

As described in detail above, the present invention has the effect of obtaining an even picture quality for the entire picture by predicting and assigning the amount of bits to be used for the actual variable rate coding based on the picture quality evaluation of the resultant image of the fixed rate coding process. .

Claims (5)

A first step of encoding the fixed amount of bits generated per unit time with respect to the input image at a predetermined value; A second step of comparing / evaluating and storing an image quality of the reconstructed frame corresponding to the input image while the fixed rate encoding of the first step is performed; A third step of confirming that the fixed rate encoding for the entire image is finished and calculating and storing a target bit amount for each unit time based on the stored image quality evaluation value; The bit rate error is calculated based on the target bit amount for each unit time stored in the output buffer, and the bit rate controller adjusts the size of the quantization step based on the bit rate error and the complexity in the image to make the bit amount generated per unit time match the target value. A bit amount allocation method of a variable rate video encoder, characterized by four steps. The method as claimed in claim 1, further comprising: a first bit rate allocation step of dividing the entire section into relatively long time units and calculating a bit evaluation function (Bti) for each time unit; And a second bit rate allocation step of re-dividing each section into relatively short time units based on the first allocated bit amount Cti and calculating a bit evaluation function Bsi 'for each time unit. A bit rate allocation method of a variable rate video encoder. 3. The method according to claim 2, wherein the bit rate (Cti) for each section in the first bit rate allocation step is assigned a bit rate by using the following equation. BT: Bit amount for the entire predetermined picture Bti: Sum of bit evaluation functions for each frame included in interval ti The method of claim 2, wherein the bit rate (Rsi) of each section in the second bit rate allocation step is assigned a bit rate by using the following equation. Bsi ': Sum of bit evaluation functions for each frame included in the interval (si) f (·): Function to calculate the bit amount of each section from bit evaluation function 1. An image encoder of a variable rate encoding method, comprising: an image encoder for performing fixed rate or variable rate encoding on an input image; An image quality detector for evaluating the image quality of the reconstructed frame by comparing the fixed rate encoding output signal with the input image signal; According to the evaluation value output from the image quality detector, a target bit amount proportional to the complexity of the image is allocated hierarchically, and the image encoder can adjust the size of the quantization step according to the hierarchically allocated target bit amount. And a hierarchical bit rate allocator.