KR100240655B1 - Extraction method of compressed image content - Google Patents

Abstract

The present invention relates to a method for extracting content of compressed video information, which solves a problem in that algorithms and hardware for defining feature information are complicated, and that a large amount of calculation is required for extracting video information. In order to extract the image information directly from the encoded discrete cosine transform (DCT) coefficients prior to the Huffman decoding process, the content of the compressed image information can be processed faster by extracting the image information directly. Extraction methods are presented.

Description

Content Extraction Method of Compressed Image Information

The present invention relates to a method of extracting content of compressed image information, and more particularly, to directly extract content of image information without performing decoding from a coded discrete cosine transform (DCT) coefficient before Huffman decoding. The present invention relates to a method of extracting the content of compressed image information that can be obtained quickly with simple algorithm and hardware by extracting.

In general, most moving image information such as TV and video data is digitized and then compressed or stored in MPEG-1 or MPEG-2 and stored or transmitted. MPEG-1 and 2 used at this time are confirmed as a compression standard for storage transmission and communication of video and are widely used.

The conventional method of extracting the content of the image information from the stored or received video data is as follows. Since MPEG-1 and MPEG-2 basically use DCT, image information is encoded in 8 × 8 block units. Accordingly, in order to obtain average brightness information of an image in a DCT block in one image, a DC (Direct Current) coefficient having an average brightness component of a block among DCT series should be decoded. That is, after performing inverse variable length code (VLC) for parsing the bit stream from the DC bit stream of the encoded DCT, inverse Huffman coding and inverse differential pulse code modulation (Differential Pulse Code) are performed. The brightness information of the block can be obtained by calculating the difference component through the process of Modulation (hereinafter referred to as DPCM).

The content extraction method of image information has a disadvantage in that it is difficult to apply to a video processing field that increases computational amount and requires real-time image processing. That is, there is a problem in that hardware and algorithms become complicated when restoring DCT coefficients to obtain accurate feature information.

Accordingly, an object of the present invention is to provide a method of extracting content of image information which can minimize decoding process and obtain feature information of image in a short time with simple hardware and algorithm.

According to an aspect of the present invention, there is provided a method of extracting content of compressed image information by analyzing a variable length bit string to analyze a plurality of direct current coefficients having luminance signal information, and using the variable length code. Obtaining a magnitude of a DC component with respect to the luminance signal, obtaining a range of difference values of brightness intensity between blocks using the magnitude of the DC component with respect to the luminance signal, and using the range of the difference value, And extracting the information and the edge information.

BRIEF DESCRIPTION OF THE DRAWINGS Fig. 1 is a block diagram showing the position of each information in a bit string compressed with MPEG-2.

2 is a block diagram illustrating a process of obtaining a difference value using a DC coefficient.

3 (a) and 3 (b) are exemplary diagrams of a DC value distribution in a macro block shown to explain characteristics of DC component sizes in various directions.

<Explanation of symbols on the main parts of the drawing>

A: macro block B: one DCT block in the macro block

C: DC coefficient 11: DC component magnitude for luminance

12: difference value 13: AC coefficient

14: block termination

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

1 is a configuration diagram showing the position of each information in a bit string compressed with MPEG.

One macro block (A) is composed of a plurality of DCT blocks (B), there is 64 data in one DCT block (B). In addition, each unit DCT block B in the macro block A has a DC component magnitude (dct_dc_size) 11, a difference 12, and an alternating current (AC) coefficient for luminance. 13 and end of block 14. At this time, the magnitude (dct_dc_size) 11 and the difference 12 of the DC component with respect to the luminance are referred to as the DC coefficient (C). In addition, the first of the 64 data in one DCT block (B) is the DC coefficient and the remaining 63 data is composed of the AC coefficient. The number of DC coefficients C in the macro block A depends on the ratio of the luminance signal and the color difference signal. That is, the ratio of luminance signal: first color difference signal: second color difference signal is 6 when 4: 2: 0, 8 when 4: 2: 2, and 12 when 4: 4: 4. exist. However, since the brightness information of the image exists in the luminance signal, the brightness information of the image can be obtained by analyzing the first four signals of the generated DC coefficients. The DC coefficients of the luminance component are variable length coded (VLC) which allocates a variable bit length in accordance with the frequency of occurrence to reduce the average number of code bits. This is shown in Table 1 called the Huffman table.

Variable length code (VLC) DC component size for luminance (dct_dc_size) Difference range 100 0 0 0 One ± 1 One 2 ± 2 to 3 101 3 ± 4 to 7 110 4 ± 8-15 1110 5 ± 16 to 31 11110 6 ± 32 to 63 111110 7 ± 64 to 127 1111110 8 ± 128 to 255

As shown in [Table 1], it can be seen that as the difference of DC values per DCT block increases, the number of encoded bits increases. The size of the DC component (dct_dc_size) with respect to the luminance is calculated using the encoded bit string, and then the range of the difference value of the DC can be known from the size of the DC component (dct_dc_size).

2 is a block diagram illustrating a process of obtaining a difference value using a DC coefficient. Each block has 64 (8 × 8) pieces of data, and the circles indicated by the hatched in each block represent DC coefficients having information on the brightness difference from the previous block in the order of numbers shown.

As shown in Table 1, the DC component size (dct_dc_size) for luminance has a larger value as the brightness difference between blocks increases. Therefore, the brightness change of the block can be measured using the size of the DC component (dct_dc_size) with respect to the luminance obtained from parsing the bit string. That is, the four DC coefficients having the brightness component of the block have information of the brightness difference from the previous block in the order shown in FIG.

For example, if there is an edge in a macroblock composed of four blocks of luminance signal components, the difference between the DC component values of the DCT blocks perpendicular to the edges is large and dct_dc_size has a long value.

[Table 2] below shows the actual difference value when the magnitude of the DC component with respect to luminance is three. The first bit of the difference value can be used to determine the sign and to calculate successive differences. That is, if the most significant bit of the difference value is 0, it means a negative sign indicating a decrease in intensity of brightness. If the most significant bit of the difference value is 1, it means a positive sign indicating an increase in intensity of brightness.

Example when dct_dc_size = 3 Difference Actual value 0 -7 One -6 10 -5 11 -4 100 4 101 5 110 6 111 7

3 (a) and 3 (b) are exemplary diagrams of DC value distribution in a macro block shown to explain characteristics of DC component sizes in various directions.

d1 is the brightness difference between block 1 (31) and block 2 (32), d2 is the brightness difference between block 2 (32) and block 3 (33), and d3 is the brightness between block 3 (33) and block 4 (34). Indicates a difference. In the case of flat areas without difference in brightness, d1, d2, and d3 all have small values.

When there is a horizontal edge as shown in Fig. 3B, d2 has a large value, and d1 and d3 have a small value.

In the case of the vertical edge, d1, d2, and d3 all have large values, but d1 + d2 and d2 + d3 have small values. D1 + d2 + d3 has a large value and d2 has a small value when there is a 45 ° diagonal edge, and d1 + d2 + as opposed to a 45 ° diagonal edge when there is a diagonal edge of 135 °. d3 has a small value and d2 has a large value.

As described above, according to the present invention, accurate bit length information can be obtained in a short time through only parsing the bit string. Since this bit length information can reflect local and overall characteristics of the image, it can be applied to image processing applications such as image retrieval and scene change, and has an excellent effect of reducing the complexity of hardware and algorithms.

Claims (1)

Analyzing a variable bit string to interpret a plurality of direct current coefficients having luminance information; Obtaining a magnitude of a DC component of luminance information using the bit string of the variable length code; Obtaining a range of a difference value for brightness intensity between blocks by using a magnitude of a DC component of the luminance signal; And extracting brightness information and edge information of the image using the range of the difference value.

Images