JPH11275578A - Adaptive bit rate control method for video stream coding - Google Patents

Abstract

PROBLEM TO BE SOLVED: To decide a more suitable quantization parameter to an input image by deciding a quantization coefficient so as to match the application degree of a buffer and the characteristic of input video through the change of a QP bit rate table value based on a quantization coefficient (QP) and produced bit quantity and allocating an adaptive bit to each macro block. SOLUTION: When quantization is performed in an intra-mode by using an encoder, a video frame to be inputted is divided into a macro block unit and each block unit is quantized from a discrete cosine converter (DCT) 1 through a quantizer 2. This signal is encoded in an encoder 5 and after that, it is stored in a buffer 6 and is outputted to a connection network. Accompanying it, an intra QP bit rate table 4 is changed based on a QP after being encoded and produced bit quantity and the buffer 6 is added or subtracted only by difference between value outputted through a channel and value inputted through a source encoder.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an adaptive bit rate control method for video stream coding, and more particularly to a method for transmitting a video portion from a video service system via a very low-speed transmission medium such as a public telephone network or a mobile communication network. The present invention relates to a technique for improving the quality of a restored video by allocating appropriate bits to each block when performing compression.

[0002]

2. Description of the Related Art Research on compression techniques for video images for ultra-low-speed media has been actively conducted recently, and improvement of the performance of a source encoder has been mainly studied. However, obtaining high quality reconstructed video under a controlled bit rate depends on bit rate control techniques that define codes and parameters during video encoding. The bit rate control technique includes a forward control technique of allocating a bit rate in consideration of characteristics of an input image, and a rear end part of a source encoder such as a buffer state and a transmission rate. There is a backward control technique for determining a parameter of an encoder in consideration of characteristics.

A video encoder for an ultra-low-speed transmission medium must have excellent performance in terms of compression efficiency, and the encoder must also have a very low complexity. Therefore, most bit rate control techniques for video encoders for very low-speed transmission media control a bit rate by adjusting a quantization parameter (QP) in consideration of a buffer state. That is, the backward
It mainly uses control techniques.

[0004]

However, the bit rate control technique for a video encoder for a very low-speed transmission medium as described above uses one mathematical model (linear model, nonlinear model) for all images. Model, a Laplacian model, an exponential model, a Gaussian model, etc.), and bits are allocated according to the model, which has the following problems. That is, in the conventional bit rate control technique for a video encoder for a very low-speed transmission medium, when the characteristics of the input video greatly differ, the codes are partially matched, but the optimal coding for the input video is performed. There is a problem that parameters cannot be determined. Also, the backward bit rate control technique has a problem in that the buffer utilization is very low because the parameters are determined in consideration of only the relationship between the state of the buffer and the parameters of the source encoder.

SUMMARY OF THE INVENTION In view of the above-mentioned conventional problems, the present invention provides a video image compression technique for transmitting and storing data over a transmission medium having a limited transmission rate and storage space. Above, so that it is possible to obtain the optimal restored video quality, and in particular, it is more suitable for the input video by predicting not only the state of the buffer but also the bit rate generation amount corresponding to the characteristics of the input video from past facts. It is an object of the present invention to provide a technique that enables a determined quantization parameter to be determined.

[0006]

In order to achieve the above object, an adaptive bit rate control method for video stream coding according to the present invention provides a bit rate item in a quantization coefficient bit rate table. Storing a bit amount generated by an encoder according to a quantization coefficient value in a quantity item to create a quantization coefficient bit rate table; and dividing an input video frame into macroblock units; When quantizing the macroblock by unit, quantizing using a quantization coefficient in a quantization coefficient bit rate table according to the current buffer state, and encoding the quantized data, Changing the quantization coefficient bit rate table value based on the quantization coefficient and the generated bit amount;
Deciding a quantization coefficient to match the utilization of a buffer and characteristics of an input image through changing the quantization coefficient bit rate table value, and allocating a bit adapted to each macroblock. Features.

In the adaptive bit rate control method for video stream coding according to the present invention, the step of performing quantization using the quantization coefficient includes checking a current buffer state at the time of quantization. Determining a quantization coefficient for maximizing the use of a buffer provided in advance in a current state, and further performing quantization.

[0008]

Embodiments of the present invention will be described below in detail with reference to the drawings. FIG. 1 is a block diagram illustrating a schematic configuration of an encoder to which the bit rate control method according to the first embodiment of the present invention is applied.

A major problem of the bit rate control method for encoding video images for an ultra-low speed transmission medium is that, as mentioned above, the actual buffer size is limited as a buffer overflow, so that data loss and data loss may occur. This is to cause a large delay. Therefore, when accurately predicting the amount of bits generated by a given quantization coefficient for each macroblock during source encoding, despite the fact that the buffer is efficiently utilized and the buffer overflow does not occur, The quality of the restored video image can be improved.

In this embodiment, from such a viewpoint, a bit rate control method for determining a quantization coefficient in consideration of an input video image and a state of a buffer is provided, and a generated bit rate is determined by a given quantization coefficient. If you make a mistake,
A bit rate generated by using a block delimiter for determining whether to perform encoding for preventing a buffer overflow is controlled.

In the bit rate control method according to the present embodiment, the optimum quantization coefficient is obtained by using the history of the relation between the quantization coefficient and the bit rate generated from the source encoder using the corresponding quantization coefficient. (Quantization Property) Determine QP. This quantization coefficient QP is determined based on the following equation (1).

(Equation 1)

In the above equation (1), QP (n) is a quantization parameter for the n-th macroblock, and CP (n)
-1) indicates the amount of buffered data stream in the (n-1) th macroblock. Also,

(Equation 2) Is the bit rate transmitted over the channel for each macroblock, BS indicates the actual buffer size, μ indicates the buffer usage parameter, and f (QP) is the quantization coefficient bit rate table ( QP bit rate table) and indicates a function for the bit amount. B (n) means the bit amount actually generated when the nth macroblock is encoded by the source encoder.

At this time, the QP bit rate table of the equation (1) is constituted by a pair of a quantization coefficient and a bit amount (Amount of Bits) as shown in Table 1 below.

[0014]

[Table 1]

In the above table, the bit amount item is converted according to the value of the recently used quantization coefficient and the bit amount generated by the source encoder at that time. And
The quantization coefficient value (QP value) is initially determined. For example, in H.263, the quantization coefficient value can be a value from 1 to 31 as shown in Table 1.

For each quantized coefficient, various methods can be used for changing the bit amount item value of the table. For example, an average value or an intermediate value of bit rates generated for the same quantization coefficient used many times recently can be selected to change the bit amount item of the corresponding table.

In the present embodiment, a method of averaging the bit amounts generated with the same quantization coefficient value up to the most recently used number of 5 times is used. Such a table of the present embodiment considers the buffer state as well as the characteristics of the input image, in contrast to the above-described technique based only on the conventional mathematical model in which only the state of the buffer is viewed. As a result, an appropriate quantization coefficient can be determined. In other words, in this embodiment, since the QP bit rate table is changed in consideration of the characteristics of the input video image, the quantization coefficient can be determined by reflecting the characteristics of the input image.

Further, the QP bit rate table is applied differently between the case of the intra mode and the case of the inter mode. However, if a macroblock is coded between the intra mode and the inter mode, , A large difference occurs in the bit rate generation amount, and the displaced frame difference (DFD) and the macroblock of the original video have different statistical characteristics from each other. Define and use. Of course, the bit amount value of the table is determined based on the value applied to each mode.

The above equation (1) indicates that the quantization coefficient for the current macroblock is a relation between the quantization coefficient and the bit rate when the quantization coefficient is used, and the usage rate of the buffer having a value from 0 to 1. It is shown that it is determined in consideration of the above. Therefore, the closer the buffer usage rate is to 1, the lower the source encoder has a lower quantization coefficient value and can encode the input video image with good quality.
The probability that a buffer overflow will occur is relatively high.

Next, based on the above-described theory, the process of performing quantization by applying the adaptive bit rate control method for video stream coding according to the present embodiment will be described.
A description will be given based on FIG.

When quantization is performed on the intra mode using the encoder shown in FIG. 1, when a video frame (video stream) is input, this video frame is divided into macroblock units, After a discrete cosine transform is performed through a cosine transformer (DCT) (1), the data is quantized through a quantizer (2). At this time, the quantization coefficient (QP) output from the rate controller (3) is input to the quantizer (2), and the quantization coefficient (Q
P) is determined based on the current buffer state and the value in the QP rate table (4) for the intra mode, but is determined as a quantization coefficient for maximizing the predetermined buffer utilization. .

When the quantized coefficients obtained as described above are input and quantized, the quantized signal is input to an encoder (5), where it is encoded and then buffered ( 6) and output to the network. Accordingly, the intra QP bit rate table is changed on the basis of the quantized coefficient after the encoding and the generated bit amount, and the buffer (6) stores the value output through the channel and the source encoding. Is adjusted by the difference from the value input through the input device.

Data input after the next frame passes through an inverse quantizer (7) and a discrete cosine inverse transformer (IDCT) (8) in units of macroblocks, and is then stored in a memory (9) in units of frames. Is stored. The data stored in the memory 9 is input to a motion estimator and compensator 10 to perform motion compensation, and then an intra / inter mode is determined in consideration of a difference between macroblocks. Then
The QP bit rate table suitable for the mode is quantized into coefficients converted in block units, and the table is modified as described above using the implementation result. At that time, a table correction method for averaging recently used values of 5 degrees was used.

The preferred embodiments of the present invention have been disclosed for the purpose of illustration, and those skilled in the art will be able to make various modifications, changes, additions, etc. within the spirit and scope of the present invention. Such modifications, changes and the like should be regarded as belonging to the appended claims.

[0025]

As described above, the present invention is more adaptable to an input image than a bit rate controller based on a conventional mathematical model. That is, the existing model generally provides a model between the buffer and the quantization coefficient without considering the characteristics of the input video, and determines and uses the quantization coefficient accordingly. The control method uses the quantization coefficient (QP) bit rate table based on the fact that the bit rate generated depending on the activity of the input video as well as the state of the buffer. It has the advantage of using it.

Furthermore, although the buffer utilization is very important for video image compression, the restored image quality shows a large difference particularly in accordance with the buffer utilization. Accordingly, the present invention has an effect that high quality video can be obtained by maintaining the maximum possible buffer utilization.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a schematic configuration of an encoder to which a bit rate control method according to a first embodiment of the present invention is applied.

[Explanation of symbols]

REFERENCE SIGNS LIST 1 discrete cosine transformer (DCT) 2 quantizer 3 rate controller 4 quantization coefficient bit rate table (QP bit rate table) 5 encoder 6 buffer 7 inverse quantizer 8 discrete cosine inverse transformer (IDCT) 9 Memory 10 Motion estimation and compensator

Claims (2)

[Claims] 1. A step of providing a bit amount item in a quantization coefficient bit rate table and storing a bit amount generated by an encoder in accordance with the quantization coefficient value in the bit amount item to create a quantization coefficient bit rate table. Dividing the input video frame into macroblocks, and using quantization coefficients in a quantization coefficient bit rate table according to a current buffer state when quantizing the divided macroblocks in units. After encoding the quantized data, changing the quantization coefficient bit rate table value based on a quantization coefficient and a generated bit amount; and Determining the quantization coefficient to match the utilization of the buffer and the characteristics of the input video through changing the quantization coefficient bit rate table value; Wherein, wherein the allocating bits to accommodate the respective macroblock adaptive bit rate control method for the video stream coding. 2. The step of performing quantization using the quantization coefficient, wherein a current buffer state is checked at the time of quantization, and a quantization is performed so as to make maximum use of a predetermined buffer utilization degree in the current state. The adaptive bit rate control method for video stream coding according to claim 1, wherein the coefficient is determined and further quantized.