KR0128861B1 - Method for changing quantizer step size - Google Patents

Abstract

Image compression method for improving a convergence speed toward a steady state varies a quantization level in correspondence with a screen variation state when quantizing a video signal. The image compression method detects a screen variation generation when a variation of a displacement frame difference(DFD) is beyond a threshold value of FD and the macro block is positioned in an intra area; counts the number of macro blocks having a varied screen, detects a screen variation generation when the number of macro blocks are beyond a threshold value of the screen variation macro blocks, and detects a generation of a screen variation; greatly quantizes a high frequency component among conversion coefficients via DCT(240), reduces a high frequency error caused by an inferior motion compensation, and enhances a convergence speed toward a steady state.

Description

A method of changing the quantization level of an image encoding device

1 is a block diagram showing a general image encoder.

2 is a block diagram illustrating an image encoder according to the present invention.

Figure 3 (a) is a diagram illustrating a method for determining the MC / non MC applied to the present invention (b) is a diagram illustrating a method for determining Intra / Non Intra.

4 is a diagram illustrating a quantization structure of a general Intra macro block.

* Explanation of symbols for main parts of the drawings

10, 210: motion extractor, 20, 220: frame storage and predictor

30, 230: Subtractor 40, 240: DCT

50, 250: quantizer 60, 260: inverse quantizer

70, 270: IDCT 80, 280: Adder

410: screen change detector 420: device when converting

The present invention relates to a video compression transmission method, and more particularly, to a method of changing a quantization level of a video encoding apparatus in which a quantization level is changed in response to a change state of a screen at the time of quantization of a video signal.

In general, a video encoding apparatus for video compression transmission is a device for providing video services such as a video phone, a video conference, and the like, and a signal processing technology of a next generation digital high definition television (HDTV). An apparatus for encoding image data in order to transmit image data to a channel having a limited transmission band so as to efficiently represent image information while minimizing distortion of a signal.

1 is a block diagram of a general video encoder for compressing and encoding video data. Such a general image encoder includes a motion extractor 10, a frame storage and predictor 20, a subtractor 30, a discrete cosine transform (DCT) 40, a quantizer 50, an inverse quantizer 60, and inverse discrete Cosine transform (IDCT) 70, adder 80, variable length coder (VLC) 100, MUX 110, buffer 120, and regulator 130.

The motion extractor 10 inputs a frame of a digital video signal to extract motion information of the frame, that is, a motion vector and a picture mode. The frame storage and predictor 20 predicts a frame adjusted in the encoding order input to the motion extractor 20 by using the motion information extracted from the motion extractor 20 and the previous frame stored in advance. The subtractor 30 removes redundancy between frames by subtracting the data output from the motion extractor 10 to data predicted by the frame storage and predictor 20 before transform encoding the data. The DCT 40 inputs the prediction error from the subtractor 30 to compress the overlap in the spatial direction.

The quantizer 50 inputs the extruded signal output from the DCT 40 and allocates it by a preset quantization level. The VLC 100 converts a code word having a high occurrence frequency into a short length code word according to a frequency of occurrence of each code word included in the quantized transform coefficient through the quantizer 50, and a code word having a low occurrence frequency. The total data amount is reduced by displaying a long code word. The MUX 110 transfers the compressed data through the VLC 100 to the buffer 120 according to the motion information provided by the motion extractor 10. The buffer 120 generates a control signal capable of controlling the amount of data output to the MUX 110 while temporarily accommodating the data so that the data output through the MUX 110 can be transmitted through a channel having a predetermined bandwidth. The controller 130 keeps the data amount constant by adjusting the quantization level of the quantizer 50 so as not to exhaust or overflow the data in the buffer 120.

Meanwhile, the inverse quantizer 60 and the IDCT 70 restore the signal output from the quantizer 50 as a reverse process of the quantizer 50 and the DCT 40 to a difference signal before transform encoding. The adder 80 adds the signal passed through the IDCT 70 and the signal passing through the frame storage and predictor 20 and transfers it to the block 20 again for the above-described operation process when the next frame is input.

In the general image encoder configured as described above, the image may be rapidly changed due to perspective movement, panning, and the like.

This screen change causes a sharp deterioration in image quality, and it takes a considerable time to return the deteriorated image quality to a normal state.

That is, a sudden change in the screen makes it difficult to extract motion information, which may cause a large error. In addition, due to a sudden screen change, the macro block may be identified as an intra mode, thereby causing a rapid increase in the amount of data. The sudden increase in data causes an increase in the quantization level, and deterioration of the screen occurs due to the increase in the quantization level.

The deterioration of the image quality due to the rise of the quantization level causes the deterioration of the accurate motion information, which leads to the deterioration of the screen. This repetition process usually returns to normal when about 10 frames have elapsed.

SUMMARY OF THE INVENTION The present invention has been made to solve such a problem, and an object of the present invention is to detect a sudden change in a screen, and when a sudden change in a screen is detected, by encoding a quantization level for high frequency components among DCT transform coefficients. The present invention provides a method of changing a quantization level of a video encoding apparatus that can reduce the amount of data.

A method of changing a quantization level of an image encoding apparatus according to the present invention is a video compression method in which a quantization process is performed after discrete cosine transforming of an image signal in units of macro blocks, wherein the macro block is determined from a motion compensation / motion compensation determination curve of an image signal. The image signal is greater than or equal to the threshold value T1 of the frame difference FD, and the image signal is present in an area greater than or equal to the threshold value T2 of the displacement frame difference DFD in the inter / intra determination curve, thereby applying a screen change macroblock. Determining; Determining whether the macro block determined as the screen change macro block is greater than or equal to a predetermined threshold value T3; And increasing the quantization level of the high frequency component of the discrete cosine transform coefficient if the macro block determined as the screen change macro block is equal to or greater than a predetermined threshold value T3.

Hereinafter, described in detail with reference to the accompanying drawings of the present invention.

2 is a block diagram illustrating an improved image encoder for implementing an image compression method for improving a convergence speed to a steady state for a screen change according to the present invention.

The improved image encoder consists of blocks 200 and 400. Here, the block 200 is the same conventional technique as described above, the motion extractor 210, the frame storage and predictor 220, the subtractor 230, the DCT 240, the quantizer 250, the inverse quantizer 260 ), IDCT 270, adder 280, buffer 290, and adjuster 300, block 400 includes screen change detector 410 and a convergence device to implement the present invention. 420 and connected to a block 200 of the prior art.

Accordingly, the block 200 of the prior art is a general technology that can be easily understood by those skilled in the art, and since the functions and the entire operation of each block have been described above, a description thereof will be omitted herein. Only the bay will be explained in detail.

Block 400 provides a method of detecting a change in a screen when changing the screen as described above.

For example, in case of sudden screen change in the image sequence, the difference between the two frames is large because the correlation between two frames is low and the motion extraction between frames is poor. The difference signal with the frame, that is, the displacement frame difference (DFD), is also large.

Accordingly, the present invention describes a method in which the screen change detector 410 detects a screen change by combining the MC / non MC determination method and the intra / non intra determination method proposed in the CCITT Recommendation.

FIG. 3 (a) shows the aforementioned MC / non-MC determination method, where Z is the absolute difference of FD and M is the absolute difference of DFD.

In Fig. 3 (b), an intra / non-intra determination method is shown, where Vard represents a DFD change of a macro block, and Varc represents a change of a macro block.

First, the poor motion estimation places each macro block on the upper right side in FIG. 3 (a), and each macro block on the lower right side in FIG. 3 (b) because the change of the DFD signal makes it larger than the change of the original signal. Place them.

The following equation (1) is a formula for determining the screen change for each macro block.

SCMB = (FD ≥ T1, non MC) and (Vard ≥ T2, intra) (1)

Here, SCMB represents a screen change macro block, T1 represents a threshold value of the FD, and T2 represents a threshold value of the DFD change.

Equation (1) is a formula for determining a screen change for each macro block. In the MC / non MC determination method, the FD is greater than or equal to T1 and the macro block is located in the non-MC region. The macro block is located in the intra area due to T2 or more, and the macro block is determined as a screen change macro block.

The following equation (2) is a formula for counting the number of screen changed macroblocks already determined in the above-described equation (1) and determining that the screen change has occurred with this count value.

Screen change = number of SCMB ≥ T3 (2)

Here, T3 represents a threshold value of the number of screen change macroblocks for determining screen change determination.

As can be seen from Equation (2), when the number of screen change macroblocks determined in Equation (1) is greater than or equal to T3 in the frame, it is determined that a screen change such as perspective movement or plane movement has occurred.

When the screen change detector 410 detects that the screen change has occurred in the above-described manner, the conversion device 420 receives the change and operates to improve the speed of convergence to the steady state encoding.

In the present invention, the following method is proposed to reduce the data congestion phenomenon when changing the screen in the above-described problem of the general image encoder.

That is, high frequency components among the transform coefficients transformed through the DCT 240 upon screen change are quantized to be much coarser than the prior quantization technique similarly to the quantization structure of the intra macroblock shown in FIG. By increasing the quantization level, high frequency errors caused by poor motion compensation are reduced. Accordingly, the prediction error is reduced by reducing the data reconstruction error in the encoder, and this reduction of the prediction error reduces the data congestion to the buffer 290. Therefore, the convergence speed to a steady state can be much faster than the previous coding method.

The following equation (3) is a formula for determining that the normal state has been returned by the above-described method.

Steady state = frequency of quantization levels with (intra-blocks in frame ≤ T4) and (T5 ≤ Q ≤ T6) ≥ T7 (3)

Here, T4 represents the number of macro blocks selected as the intra region, T5 and T6 represent quantization levels returned from the controller, and T7 represents a range of frequencies of the quantization levels.

That is, equation (3) indicates that the determination of the steady state is determined by the number of intra blocks and the quantization level fed back from the buffer. After the normal state, encoding is continued by the general method.

As described above, the present invention is a quantization of the result of the discrete cosine transform of the motion-compensated prediction error after detecting the screen change, and then roughly quantizes the high-frequency component to generate a high frequency error caused by the poor motion compensation. Compression, and reduce the data recovery error in the encoder to reduce the prediction error, thereby reducing the data congestion to the buffer has a great advantage that can perform the convergence to a steady state much faster.

Claims (2)

In an image compression method for performing quantization after discrete cosine transform of a video signal in units of macro blocks, the video signal of the macro block is present in a motion compensation region in a motion compensation / motion compensation determination curve of the video signal. Determining whether a video change macroblock exists in an intra region in an intra determination curve; Determining whether the number of the macroblocks determined as the screen change macroblocks is greater than or equal to a predetermined threshold value T3 per frame; And raising the quantization level for the high frequency component of the discrete cosine transform coefficients if the number of the macroblocks determined as the screen change macroblocks is greater than or equal to a predetermined threshold value T3. . 2. The quantization level according to claim 1, wherein the number of intra blocks in a frame is smaller than a predetermined threshold T4 (number of macro blocks selected as intra regions) after the quantization level of the discrete cosine transform coefficient is increased for the high frequency component. Returning to the normal quantization level when the frequency present between the predetermined thresholds (T5 and T6) is greater than the predetermined threshold (T7).