KR100644570B1 - Bitrate control method and apparatus of image encoder and encoding method therefor - Google Patents

Abstract

Disclosed are a method and apparatus for controlling a bitrate using information generated by an encoder and an encoding method using the same. A bitrate control method for controlling a quantization step of an encoded image to make a constant bitrate output from a buffer, comprising: a quantization model indicating a correlation between a bitrate and an energy of an image encoded in a predetermined range of an estimated value when encoding an image A plurality of quantization models, a process of selecting one of the quantization models set in the process by referring to an estimated value generated during encoding, a bit rate output from the buffer with reference to the quantization model selected in the process and the available capacity of the output buffer Determining the quantization step to be constant. The bitrate control method according to the present invention has an effect of reducing the error between the number of bits actually generated and the number of prediction bits by changing the modeling according to the video encoding information.

Description

Bitrate control method and apparatus of an image encoder and an encoding method suitable for the same {{Bitrate control method and apparatus of image encoder and encoding method therefor}

1 is a graph for schematically showing a quantization model used in a conventional encoding method.

Figure 2 is shown to schematically show a method of determining a quantization model in the bitrate control method according to the present invention.

3 is a block diagram showing the configuration of an image encoder including a bitrate control apparatus according to the present invention.

The present invention relates to a video encoding method and apparatus, and more particularly, to a method and apparatus for controlling a bit rate using information generated by an encoder and an encoding method using the same.

In general, the purpose of controlling the bit rate in video encoding is to maintain a constant transmission amount. That is, the quantization size is determined in order to prevent overflow or underflow of the encoded data.

To this end, the structure for controlling the bit rate should be simple and real-time operation should be possible, and at the same time, accurate judgment should be made.

Therefore, when determining the quantization size, which is the most important factor in bit rate control, the energy, given rate, and quantization size of the differential image, which is the difference image between the previous frame and the current frame, are modeled.

However, the modeling approach has a disadvantage in that it is difficult to calculate an accurate value because an error occurs in the value of the encoding bit using only energy.

An object of the present invention is to provide an improved bitrate control method for setting an appropriate number of bit estimation model and selecting one of the models set using encoding information corresponding to a macroblock being encoded to reduce an error value.

Another object of the present invention is to provide an encoding method using the above bit rate control method.

Another object of the present invention is to provide an apparatus suitable for the bitrate control method.

In order to achieve the above object, the present invention provides a bitrate control method for controlling a quantization step of an encoded video to make a constant bitrate output from a buffer.

Setting a plurality of quantization models representing a correlation between energy and bit rate of an image encoded within a predetermined range of an estimated value when encoding the image;

Selecting one of the quantization models set in the above process by referring to an estimated value generated during encoding;

And a step of determining a quantization step to make the bit rate output from the buffer constant by referring to the selected quantization model and the available capacity of the output buffer.

In order to achieve the above object, the present invention provides a DCT encoder for DCT encoding a differential image, a quantizer for quantizing a DCT coefficient resulting from the DCT encoder, and a buffer for buffering and outputting the bitrate resulting from the quantizer. In the image encoding apparatus for encoding the difference image between the original image and the motion estimation image,

An energy calculator configured to calculate energy of the difference image;

A model selection unit for selecting one of quantization models representing a correlation between energy and bit rate of a plurality of encoded images set within a range of estimated values of the image;

And a model lookup table for controlling the quantization step of the quantizer according to the quantization value generated in the quantization model selected by the model selection unit, the energy of the motion vector and the difference image, and the available capacity of the buffer. .

Hereinafter, the configuration and operation of the present invention will be described in detail with reference to the accompanying drawings.

1 is a conceptual diagram illustrating a bitrate control method by general modeling.

In FIG. 1, the horizontal axis represents energy of a macro block to be encoded, and the vertical axis represents bit rate. Each graph also shows the relationship between the bitrate and the energy of the macroblock according to the given quantization steps Q1, Q2 and Q3.

The energy of the macroblock used in the graph of FIG. 1 is an MSE (Mean Squar Error) or a variation of the macroblock to be encoded in the difference image.

An image encoding method using a bitrate control method by general modeling is performed as follows.

1) A quantization model as shown in FIG. 1 is set.

2) Enter the number of bits available in the buffer.

3) Calculate the energy of the corresponding macroblock.

4) The quantization size is determined using the quantization model shown in FIG. 1 to estimate the number of bits needed to encode the corresponding macroblock. That is, the bitrates resulting from quantizing a macroblock having a given energy into various quantization steps Q1, Q2, and Q3 are estimated, and a quantization step for calculating a bitrate suitable for the available capacity of the buffer is determined. .

5) Encode the corresponding macroblock with the determined quantization size.

6) Repeat steps 1-5 to encode the whole image.

In the above encoding method, since the number of encoded bits is estimated using only the energy of the macro block to be encoded, there may be an error in the value thereof. In contrast, a margin should be considered in the size of the buffer. .

In order to solve this problem, the present invention is characterized by reducing the control error by setting an appropriate bit number estimation model and selecting one of the set models using the encoding information corresponding to the macroblock being encoded.

In the bitrate control method according to the present invention, the quantization size Q may be represented in the form of a function as follows.

Pi, i = 1, ......, N: Predictors

Mk, k = 1, ......, M: Models

Q = f (Mk / Pi)

For example, if the magnitude and variance of the motion vector MV are used as the estimated value for the image encoding information, the first predictor P1 = MV. For example, if M1, M2, and M3 are set for a predetermined range of the magnitude of the motion vector, one of M1, M2, and M3 is selected with reference to the motion vector. On the other hand, if these models are set up using the relationship between the variance and the rate, the variance value is obtained, the desired rate is input, and then the quantization value is obtained. At this time, another control signal can be generated using this quantization value Q.

2 is a block diagram showing the configuration of an image encoder including a bitrate control apparatus according to the present invention. The apparatus shown in FIG. 2 inputs a blocked differential image signal and outputs a bit rate encoded in units of blocks. In addition, the quantization step of the DCT coefficient is controlled by referring to the available capacity of the buffer and the preset quantization model. Specifically, when the available capacity of the buffer is small, the quantization step is increased to reduce the amount of bits generated.

In the apparatus shown in FIG. 2, the input image is pre-processed in blocks by noise filtering by a temporal and spatial filter in the preprocessor 210, and the blocked encoded image is DCT coded for each block in the DCT 212. The DCT coefficients are variable quantized in the quantizer 214. The bitrate obtained as a result of the quantization is output through the buffer 216.

The lookup table 228 refers to the quantizer 214 by referring to the available capacity of the buffer 216, an estimated value (for example, a motion vector value, a DCT value, and an image data value) that is image encoding information, an energy of a block, and a quantization model stored therein. Control the quantization step. In addition, the control signal Vj = f (Q), j = 1, ..., L, it is possible to adjust the strength of the pre-processing of the preprocessor 210.

Estimates and energy needed to determine the quantization step are obtained from model selector 226 and energy calculator 224, respectively.

The model selector 226 includes a plurality of quantization models that are set in a predetermined range of estimated values by referring to an estimated value generated during encoding, for example, a motion vector value, DC and AC values of the DCT 212, or an image data value. Choose one.

Inverse quantizer 218, inverse DCT 220, and memory 222 are used to obtain previous and current frames and generate a motion vector (MV) therein.

3 is a diagram illustrating a bitrate control method according to the present invention.

Predictor 1-Predictor N 310-320 are encoding information. For example, Predictor 1 310 may be a motion vector value, Predictor 2 may be a DC value in units of blocks, and Predictor 3 may be a frequency component in units of blocks. have. The model selector 330 selects model 1-model M 340-370 by referring to any predictor value, for example, a motion vector value, among the predictor 1-predictor N 310-320. In another embodiment, the predictor 1-predictor N may be combined with each other by applying different encoding information, and models corresponding thereto may be selected. Model 1-Model M (340-370) are quantization models representing the relationship between the bit rate and the energy of the corresponding block divided into a plurality of estimates, that is, a certain range of motion vector magnitudes. For example, Model 1-Model M (340-370) are classified according to the motion vector size, and arbitrarily set Model 1 for the largest motion vector and Model M for the smallest motion vector. Accordingly, appropriate models are selected according to the magnitude of the motion vector value generated in any one predictor 310. The selected model also generates a control signal accordingly.

The bitrate control method according to the present invention sets M quantization models (model 1 to model M) for a predetermined range of estimated sizes using encoding information, and uses the selected one of them to encode the difference image. Estimate For example, the magnitude value of the motion vector in the encoded information may be set as an estimated value. The reason for using the magnitude of the motion vector as the front-end step is that the larger the motion vector value is, the more motion the image is, and thus, it is difficult to accurately estimate the motion. This is because it tends to affect the difference image.

Therefore, by classifying macroblocks with many errors and macroblocks with relatively few errors beforehand, more accurate modeling and prediction of the number of bits to be coded are possible.

Modeling generates experiments on the relationship between the number of bits and energy generated by adjusting the quantization size for many test images. As a result, the quantization magnitude is controlled using the above probability, A Priori.

Looking at the case of using a motion vector as an estimated value, the size of the motion vector can be determined as follows.

The motion vector MV is called (MVx, MVy), and the magnitude of the motion vector is represented by | MV | = | MVx | + | MVy |.

Here, MVx and MVy represent the magnitudes of the motion vectors in the x- and y-axis directions, respectively. Depending on the magnitude of the vector value and the correlation between the models, it can be divided into n vector size regions. That is, it can be represented by {MVR1, MVR2, .... MVRn-1, MVRn}.

The models are constructed similarly to the modeling shown in FIG. 1. That is, the vertical axis represents the bitrate and the horizontal axis represents the magnitude of energy, and they have a single relationship for all Q {Q = min, ..., max}. For example, in H.263, the min value is 1 and the max value is 31. The number of models is possible with M models when using a normalized model.

The estimated number of bits Re (q) is predicted in the following form.

Q = 0, step = 16;

do {

   Re (q) = EstimateBits (Q);

   step / = 2;

   if (Re (Q)> number_of_target_bits)

      Q + = step;

      else

     Q-= step;

   } while (step> 0);

After executing this iterative statement, the final Q value can be estimated. In the case of H.263, the final result is obtained after five iterations.

According to the present invention, an encoding method using the bitrate control method in which an estimate is determined as a motion vector is performed as follows.

1) Set the estimate to the motion vector.

2) A plurality of quantization models representing the correlation between the energy and the bitrate of the encoded image within the range of the motion vector of the encoded image are set.

3) The difference image between the original image and the motion compensated image is calculated.

4) The motion vector value calculated during the motion estimation is read and the vector size region is determined according to the size.

5) A quantization model corresponding to a predetermined vector size region is selected from the plurality of quantization models of 2).

6) Find the difference image energy.

7) Accept the available capacity from the buffer.

8) Determine the quantization step from the difference image energy and the selected model.

9) Quantize the DCT coefficients by applying the calculated quantization magnitude.

As described above, the bitrate control method according to the present invention sets appropriate and various bit number estimation models and selects one of models set by using encoding information corresponding to a macroblock to determine the number of bits actually generated and the number of predicted bits. It has the effect of reducing the error.

In addition, accurate prediction of the bit generation amount is easy to optimize the size of the buffer, and the size of the buffer can be reduced to reduce the delay loss caused by the buffer.

In addition, since it is a buffer control method by modeling, there is an advantage that control is possible in real time.

Claims (7)

A bitrate control method for controlling a quantization step of an encoded video to make a constant bitrate output from a buffer, Subdividing and setting a plurality of quantization models representing a correlation between the energy of the image to be encoded and a bit rate in a predetermined range of the magnitude of the estimated value at the time of encoding the image; Combining at least one estimated value generated during encoding and selecting one of the quantization models set in the above process according to the magnitude of the estimated value; And determining a quantization step of making the bit rate output from the buffer constant by referring to the selected quantization model and the available capacity of the output buffer. The method of claim 1, wherein the estimated value is a motion vector value. The bit rate control method according to claim 1, wherein the estimated values are DC and AC values in units of blocks. The method of claim 1, wherein the quantization model setting process is performed. Determining one of the n classes according to the magnitude of the motion vector; And And selecting a class corresponding to the determined class among n preset quantization models. An image encoding method for encoding a difference image between an original image and a motion estimated image, Obtaining a difference image between the original image and the motion estimation image; Subdividing and setting a plurality of quantization models representing a correlation between energy and bit rate of an encoded image within a range of estimated magnitudes of the encoded image; Combining at least one estimated value generated at the time of encoding and selecting one of the quantization models set in the process according to the magnitude of the estimated value ; Determining a quantization step of making the bitrate output from the buffer constant by referring to the selected quantization model, the energy of the difference image, and the available capacity of the output buffer; And And quantizing the difference image by applying the determined quantization step. The method of claim 5, wherein the quantization model used in the quantization model setting process represents a correlation between a bitrate normalized by the energy of the encoded image and the magnitude of the motion vector. A DCT encoder for DCT encoding the differential image, a quantizer for quantizing the DCT coefficients resulting from the DCT encoder, and a buffer for buffering and outputting the bit rate resulting from the quantizer; An image encoding apparatus for encoding a difference image of An energy calculator configured to calculate energy of the difference image; A model selection unit that selects one of quantization models representing a correlation between energy and bitrate of a plurality of encoded images set within a size range of an estimated value of the image as the size of the estimated value by combining at least one or more estimated values; And a model lookup table for controlling the quantization step of the quantizer according to the quantization value generated in the quantization model selected by the model selection unit, the energy of the motion vector and the difference image, and the available capacity of the buffer.

Images