KR890003612Y1 - Imaging signal coding system - Google Patents

Abstract

No content.

Description

Video signal compression coding device

1 is a schematic configuration diagram of the device of the present invention.

2 is a conversion coefficient statistical table according to the present invention device, b, c is a configuration diagram of the conversion coefficient compressor and the expander according to the present design device, d is a conversion quantization characteristic diagram by the present invention device.

3 is a schematic diagram of a conventional encoder.

The present invention relates to a video signal compression encoding apparatus in a video signal transmission apparatus, and more particularly, to a video signal compression incubation apparatus which encodes a video signal having a moving part as statistical and visual characteristics to enable low speed transmission.

Conventionally, as shown in FIG. 3, an analog / digital converted and filtered video signal is inputted to the block configurator 1, and the video signal inputted in the scanning order is configured in units of blocks. At this time, the predictor 8 reads out the image signal value of one block of the same size at the same position spatially in all frames, calculates the difference in the subtractor 2, and uses the difference as a prediction error value.

These prediction errors are transform-coded by the transform encoder 3 and then quantized by the quantizer 4 and transmitted.

On the other hand, in order to encode the next frame, the inverse quantizer 5, the inverse transform encoder 6, the inverse process of the transform encoder 3 and the quantizer 4, adds the prediction value and the prediction error value in the summer 7, and then stores the memory. It is a structure to store in (8).

However, there may be digital filters for noise reduction between the analog / digital converter 1 and the subtractor 2.

This is a disadvantage that the compression of the amount of information is not enough for low speed transmission is impossible to transmit at low speed.

This paper adopts block-based moving compensation and variable-length coding using statistical properties of video signals and selects transform coefficients using visual characteristics so that low-speed transmission can be achieved without significantly degrading image quality. An object of the present invention is to provide an apparatus for compressing and encoding a video signal. Hereinafter, the effect of the construction of the present invention will be described with reference to the accompanying drawings.

However, the digital filters for noise reduction are omitted and only the transmitting side is mentioned.

The receiving side may go through the reverse process of the transmitting side.

Referring to FIG. 1, the structure of the present invention includes a block configurator 1, a subtractor 2, a transform encoder 3, a quantum 5, an inverse quantum 11, a transform coefficient expander 12, and an inverse transform. In the video signal compression encoding apparatus including the encoder 13, the adder 14, and the predictor 15, the output of the transform encoder 3 is connected to the quantizer 5 through the transform coefficient compressor 4, The output of the quantizer 5 connects the variable length coder 6 to which the Huffman code table 7 is connected to the multiplexer 8, and the output of the multiplexer 8 is connected to the output 18 via a buffer 9. The buffer 9 is connected to the conversion coefficient compressor 4 and the conversion coefficient expander 12 through the buffer controller 10, and the output of the summer 14 is transferred through the frame memory 16 to the moving estimator 17. In addition to connecting the output of the block estimator (1) and the output of the mobile estimator (17) to the predictor (15) and the multiplexer (8). First, the statistical characteristics of the video signal and human visual characteristics are explained. Since the video signal is a lowpass signal, not only has a significant correlation between adjacent pixel values in a frame, but also the frame rate is NTSC. At 30 per second, there is a significant correlation between frames, except when the scene changes or there is intense movement. On the other hand, the more detailed the image in the frame, the less sensitive the human eye is and the rise time of the human eye is 80 msec even between successive frame poles. Therefore, the human eye becomes spatial and temporal ( It has the effect of lowpass filtering on both directions of temporal direction.

In this paper, the relationship between the statistical characteristics and the visual characteristics of the video signal described above will be described with reference to FIG. 1. Referring to FIG. 1, one block (for example, 4x4, 8x8, etc.) is provided through the block configurator 1. The pixel values of the formed current frame and the prediction value of the previous frame of the same sized block output from the predictor 15 are calculated by the subtractor 2 to calculate a difference, that is, a prediction error.

This reduces the amount of information by eliminating redundancy or redundancy between frames by performing DPCM (Differential Pulse Code Modulation) between frames in blocks.

Here, in order to obtain a good prediction value, that is, to reduce the amount of information as much as possible, in the block configurator 1 in the irradiation area of a predetermined size among the pixel values of the previous frame stored in the frame memory 16 in the moving estimator 17 The block having the maximum correlation with the input block pixel values of the output current frame is found, and these pixel values are output to the predictor 15 as predicted values, and the positions thereof are output to the multiplexer 8 as a moving vector to a decoder (not shown). To be used to restore the screen). At this time, the motion estimator 17 finds the block of the previous frame having the maximum correlation with the input block of the current frame under the assumption that the object moves on the screen. As a measure, MSE (Mean Square Error) and MAE (Mean Absolute) Error: Absolute error mean value, and 2D (2D) algorithm or known one-time survey method (OTS: One at a tine Searcl), Roga to find the minimum mismatch with minimum calculation. You can use rhythm research methods.

Therefore, in the case of a block belonging to the part without motion, a block having the same position spatially in the previous frame is used as the prediction value, and in the case of a block belonging to the part having motion, the most similar block in the previous frame is used as the prediction value.

Therefore, the prediction error values output from the subtractor 2 are two-dimensional (2D) transformed in the frame by the transform encoder 3.

In this case, DCT (Discrete Cosine Transform) or DFT (Discrete Four Transform) can be used.

The transform coefficient compression and extension method in this paper is a combination of region sampling and critical sampling using statistical and visual characteristics.

For example, in case of applying the conversion encoding in units of (8 × 8) blocks, the conversion coefficient is E (i, j): 0 i, j Therefore, if 2 is arranged in frequency order as shown in FIG. 2, E (0,0) is a dc component (in case of DST (discontinuous sine transform) without a DC component, the block calculus is first subtracted) As described above, according to the statistical characteristics of the video signal, high frequency components are gradually increased in the FX and FY directions.

Then, the sensitivity of the high frequency component is inferior due to the characteristics of the human eye as described above. You can reduce the amount of information without damaging it.

In other words, the statistical characteristics of the transformation coefficients arranged as shown in FIG. 2 are less variance toward the fx and fy directions.

Also, the visual characteristic of this arrangement is that the higher the frequency, the lower the temporal sensitivity.

In the transform coefficient compressor 4, critical sampling is performed according to the occupancy level of the buffer 9 so as to cope with a change in the amount of information according to the amount of moving part of the screen. However, the sensitivity of the human eye to the moving part is almost irrelevant if the temporal resolution is good even if the spatial resolution falls, that is, the tolerance of the human eye to the blurring of the blur in the moving part. In this paper, even if the amount of data in one frame is compressed, it is overcome because it does not skip the frame.

Therefore, by dividing the quantized transform coefficients into three groups, the first group allocates a fixed number of bits (for example, 8 bits), and the second group reads and encodes a variable Huffman code from the Huffman code table (7). The conversion coefficient below the threshold is treated as 0 so that a continuous run length of 0 is also read out from the Huffman code table 7 and transmitted. Huffman's code table (7) is therefore for nonzero coefficient amplitude.

In this case, the transform coefficients of the first group are extremely important for screen reproduction. Therefore, the fixed codes are always transmitted as they are, and the transform coefficients of the second group are subtracted from the threshold value adaptively determined according to the occupancy level of the buffer (9). Huffman encodes and transmits.

At this time, the value below the threshold is treated as 0, the number of consecutive zeros is encoded and transmitted, and the third group of transform coefficients (high frequency components) are not transmitted.

On the other hand, as shown in FIG. 2 (a), the transform coefficient encoding order follows the arrow direction.

This is because, in general, it is possible to generalize that the dispersion coefficients of the low-frequency video signals are reduced in the order of arrows in FIG.

Compression and expansion of the transform coefficient are as shown in (a) and (b) of FIG.

In this design, the quantum 5 is a single uniform quantum.

This is because uniform quantizers have the least amount of quantization noise and do not require nonlinear characteristics by introducing variable length coding.

This quantum 5 uses a neutralized quantum such as (d) in FIG.

The variable length encoder 6 encodes the number of transform coefficients of 0 consecutive to the value of the non-zero transform coefficient.

In the present invention, a non-zero transform coefficient value and a sign table of consecutive zero transform coefficients are used.

This is because both cases can be regarded as having the same statistical properties because small values dominate.

The inverse quantum 11, the transform coefficient expander 12, and the inverse transform encoder 13 are process elements for converting data of the current frame transmitted for prediction of the next frame to pixel values.

This paper is very useful in the case of magnified images of head and shoulder scenes with fixed camera constant illumination compound movement or only lip movement without continuous movement. When the TV video signal is transmitted by A / D conversion (PCM) at 8 bpp (bits per pixel), a data rate of about 80 Mbps is required. However, in this paper, we assign 8 bits to the 3rd 12th coefficients in 8x8bloct, for example, as shown in (a) of the figure. If we allocate × 4 bits, or bits (assuming that approximately 3 bits per coefficient are allocated, it will vary depending on which image is targeted), and other sync signals, control signals, etc. To get the compression ratio.

As described above, according to this proposal, the video signal can be compressed, thereby enabling a low-speed transmission, thereby reducing the line usage cost.

Claims (1)

In a video signal compression encoding apparatus including a block configurator, a subtractor, a transform encoder, a quantizer, an inverse quantum, a transform coefficient expander, an inverse transform encoder, a summer, and a predictor, the output of the transform encoder 3 is a transform coefficient compressor ( 4) to the quantizer 5, the output of the quantizer 5 is connected to the multiplexer 8 via a variable length encoder 6 to which the Huffman code table 7 is connected, and the multiplexer 8 Is connected to the output 18 through the buffer 9, the buffer 9 is connected to the conversion coefficient compressor 4 and the conversion coefficient expander 12 through the buffer controller 10, the summer 14 The output of the mobile estimator 17 is connected to the mobile estimator 17 through the frame memory 16 and the output of the block estimator 17 is connected to the predictor 15 and the multiplexer 8. A video signal compression encoding device connected to the.