KR0124164B1 - Moving image decoding method - Google Patents

Abstract

A method for decoding a moving image signal is disclosed. The method comprises a first step of reconstructing a pixel based input image data to provide a block-by-block image data and providing a predetermined block which represents the position wherein the input image data exists; and a second step of processing an inverse quantization with the image data existing pixels. Since, a selective inverse quantization of the pixel based input image data which is variable-length decoded is possible, the efficiency of the decoding is increased.

Description

Video Decoding Method

1 is a block diagram illustrating a video decoding apparatus in which a video decoding method according to the present invention is executed.

2 is a diagram illustrating a video decoding method of the present invention.

* Explanation of symbols for main parts of the drawings

40: VLD part 41: reverse scanning part

42: inverse quantization unit 43: inverse discrete cosine transform unit

44: adder 45: motion compensation unit

46: frame memory section

The present invention relates to a video decoding method, and more particularly, to a video decoding method of improving a decoding speed by selectively dequantizing variable length coded video data.

The video decoding apparatus refers to an apparatus for decoding image data received through compression encoding by a known encoding method into original data transmitted. That is, a compression coding transmission method includes a discrete cosine transform (hereinafter referred to as DCT) method for removing spatial redundancy of an image signal and a motion compensation encoding method for removing temporal redundancy of an image signal. Hybrid methods combining these two methods are widely used.

In the hybrid encoding method, when a video signal is DCT, the DCT coefficients are quantized, the quantized DCT coefficients are encoded in run and level, and the level values are variable-length encoded and transmitted. That is, after DCT of a predetermined number (usually 8x8) pixel blocks, the DCT block is quantized by zigzag scanning, and the number of zeros from the quantized DCT coefficients until a nonzero coefficient is obtained is determined. Run) and non-zero coefficients as Level values.

Here, when the coefficient in the 0 state is continued to the end point of the block, it is represented (coded) by EOB (End Of Block). Quantized DCT coefficients encoded with run and level values are compressed again by variable length coding.

In order to decode the compression-encoded signal as described above, the process at the time of compression is reversed. That is, the variable length codes are decoded again into run and level values, and a quantization coefficient of 8x8 blocks is formed according to these level values. The quantization coefficients are inversely quantized to form DCT coefficients, and inverse DCT is performed on these DCT coefficients to decode the original video signal.

As described above, the conventional decoding apparatus performs an inverse quantization process on 8x8 blocks formed by the run and level values. However, since DCT coefficients per block occupy the majority of blocks with a coefficient of zero, there is a problem that it takes a considerable time to dequantize all the coefficients of zero.

SUMMARY OF THE INVENTION The present invention has been made to solve such a problem, and an object of the present invention is to provide a video decoding method which improves the execution speed during inverse quantization by performing inverse quantization only on non-zero quantization values.

In order to achieve this object, the present invention provides a video signal in which a predetermined intra-picture video signal is quantized after discrete cosine transform (DCT), quantization coefficients are quantized as run and level values, and run and level values are variable length coded. CLAIMS 1. A method of decoding, comprising: decoding a variable length code into said run and level values; Reconstructing the quantization coefficients on a temporary block basis according to the run and level values, and searching for positions of level values in the block; Inverse quantization of level values in the block and setting to 0 for a portion other than the level value to detect a DCT coefficient; And inversely transforming the DCT coefficients into a video signal.

Hereinafter, with reference to the accompanying drawings will be described an embodiment of the present invention;

1 is a block diagram illustrating a video decoding apparatus in which an improved video decoding method is implemented. The video decoding apparatus employed in the present invention includes a variable length decoding (VLD) unit 40 and an inverse quantization unit. 42, an Inverse Discrete Cosine Transform (IDCT) unit 43, an adder 44, a motion compensation unit 45, and a frame memory unit 46.

The VLD unit 40 decodes the compressed data into the run and level values when the compressed data is input in the form of a bit stream (run and level values encoded by the variable length encoding apparatus of the encoder) through the line 30. Is applied to the phase. In addition to the received bit stream information code, the VLD unit 40 additionally detects a quantization step size (QS), a motion vector (MV), and the like during encoding, and the quantization step size is The inverse quantization unit 42 transmits the motion vector to the motion compensation unit 45 which will be described later.

The inverse scanning unit 41 rearranges the two-dimensional quantized values in block units (8 × 8 blocks) to the line 32a according to the run and level values applied from the line 31. Is authorized. In addition, the inverse scanning section 41 of the present invention outputs position information indicating a position where non-zero quantization coefficient values exist for the run and level values on the line 32b.

Referring to FIG. 2, a block-based quantization coefficient generated by the inverse scanning unit 41 described above is shown. FIG. 2a is a diagram of quantization coefficients in an 8x8 block output on a line 32a. FIG. 2B is a diagram showing positions of nonzero quantization coefficients in FIG. 2A. That is, in FIG. 2A, for example, when the run 0 level 7, the run 4 level 2, the run 1 level-1, and the EOB data are generated through the above-described VLD unit 40, the inverse scanning unit 41 is generated. Figure 2b shows the data generated in Figure 2b shows the position in the block where the level values exist (shown as 1 in the figure, but this is used for illustration, it does not have a different meaning).

Referring back to FIG. 1, inverse quantization unit 42 is coupled to lines 32a and 32b, so that the in-block quantization coefficients from inverse scanning unit 41 described above are quantized parameters from VLD unit 40. The DCT coefficient is output on the line 33 by inverse quantization according to QS. At this time, the inverse quantization unit 42 of the present invention does not perform inverse quantization on the entire 8x8 block, but only on the level value by the position information applied from the line 32b (FIG. 2 (b)). Inverse quantization is performed only for the portion shown as 1 in the (), and is set to 0 for the portion other than the level value. Here, since a position having a non-zero level value among the quantization coefficients of the 8 × 8 block of the inverse scanning unit 41 is usually a prime number, the decoding method of the present invention is applied to all the quantization coefficients in the 8 × 8 block. It can be seen that the dequantization rate is significantly improved compared to the conventional method of inverse quantization.

The IDCT 43 is coupled to the line 33 and inversely transforms the DCT coefficient (IDCT) and outputs it on the line 34. The adder 44 is coupled to the line 34 so that when the input data is an intra frame, the adder 44 outputs directly onto the line 35 as shown, whereas the adder 44 is an inter frame. In this case, the motion vector from the VLD unit 40 is reversed in the same order as in encoding, so that the data of the previous frame stored in the frame memory unit 46 in the motion compensation unit 45 in the same manner as in encoding is correlated. The highest part of the figure is estimated, and the predicted image output on the line 36 is added and applied on the line 35.

As described above, according to the present invention, zero quantization coefficients among the variable-length decoded input quantization coefficients are dequantized only for a few non-zero quantization coefficients, that is, level values, without inverse quantization, thereby improving video decoding speed. There is a big advantage to improving the decoding performance.

Claims (2)

A method of decoding a video signal in which a predetermined intra-picture signal is quantized after a discrete cosine transform (DCT), the quantized coefficients are quantized as run and level values, and the run and level values are variable length coded. Decoding a length code into the run and level values; Reconstructing quantization coefficients in predetermined block units according to the run and level values, and searching for positions of level values in the block; Dequantizing the level values in the block, and setting the value to 0 for a portion other than the level value to detect a DCT coefficient; And inversely transforming the DCT coefficients into a video signal. The video decoding method of claim 1, wherein the predetermined block comprises 8 × 8 pixels.