KR0135364B1 - Dct block coding method and its apparatus using block - Google Patents

Abstract

The present invention relates to a method and apparatus for encoding a DCT block using a block-adaptive scan. In particular, the encoder 100 includes a scan pattern storage unit 11, a DCT converter unit 12, and a feature extractor ( 13), the scan pattern determination unit 14, the processor unit 15 and the entropy encoding unit 16, and the decoder 200 is a scan pattern storage unit 21, DCT switching unit 22, a feature extraction unit (23) The scan pattern determination unit 24, the entropy decoding unit 25, and the processor unit 26 constitute a scan pattern that reflects the characteristics of a general image as in the prior art, and the scan pattern is arbitrarily adjusted without uniformity. The efficiency can be greatly improved by determining the set of and analyzing the characteristics of each DCT block of the image and applying the appropriate scan pattern.

Description

Character information display device

1 is the original image taken by the camera.

2 is a motion compensation image of the current video from the previous video.

3 is a difference image when the current image is predicted by the motion compensation image.

4 illustrates one embodiment of scanning pattern sets adapted to the present invention.

5 is a chart showing the CDT coefficients used for feature extraction.

6 is a flowchart of an encoder for explaining the method of the present invention.

7 is a flowchart of a decoder for explaining the method of the present invention.

8 is a system configuration diagram of an encoder in the apparatus of the present invention.

9 is a system configuration diagram of a decoder in the apparatus of the present invention.

* Explanation of symbols for main parts of the drawings

11, 21: Stan pattern storage unit 12, 22: DCT converter

13, 23: feature extraction unit 14, 24: scan pattern determination unit

15, 26: processor unit 16: entropy encoding unit

25: entropy decoder 100: encoder

200: decoder

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a coding technique for compressing and reconstructing image data having a large amount of information, such as high-definition television (HDTV), that is, a quantization scanning technique of a discrete cosine transform (DCT) transform coefficient during image encoding and decoding. A method and apparatus for encoding a DCC block using a block-adaptive scan for adaptively converting a scan pattern to increase the efficiency of variable length incubation.

In general, image data compression is of great interest in various applications such as broadcasting, communication, and storage of video signals. The processing of digital signals has many advantages in data compression applications over analog signal processing, so the main focus of the image transmission field is on the encoding of digital images.

Moreover, this trend has been more noticeable with the introduction of digital HDTV system and Youngsan transmission through the network, and the standardization of the video encoding method in various applications has been steadily studied. Examples are CCITT Recommendation H.261 and MPEG.

Most of these systems encode a video signal using motion compensation and DCT (MC-DCT coder). Performing motion compensation and DCT removes temporal and spatial redundancy of video data.

In a typical MC-DCT encoder, a motion compensated prediction difference image is conceptually encoded in three steps as follows.

1. Two-Dimensional DCT of Motion Compensation Prediction Difference Image

2. Scalar quantization of DCT coefficients

3. Entropy Coding of Quantized Coefficients

Since most DCT coefficients are quantized to zero and nonzero coefficients are sparsely distributed, an efficient coding technique called Runlength Encoding is frequently used by using this property. Line length sublines, that is, consecutive zeros and subsequent nonzero values are entropy coded together.

For effective line length coding, scan patterns with small entropy and average coding length should be made. In this respect, zigzag scans are generally effective scan patterns.

Since the MC-DCT coder allocates approximately 90% of the total number of bits for encoding the quantized DCT coefficients, it is very important to indicate that the MC-DCT encoder is effective for encoding the quantized DCT coefficients.

For this purpose, some methods of analyzing the characteristics of DCT coefficients have been proposed. Most of the image encoding scan methods developed so far apply a uniform pattern to each image by defining a scan pattern that reflects the general characteristics of the image. There was a problem in that the efficiency is lowered.

Unlike the conventional method, the present invention sets a random set consisting of scan patterns, analyzes characteristics of each DCT block of an image, and applies a scan pattern suitable for the block-adaptation to improve its efficiency. The present invention provides a method and apparatus for encoding a DCT block using a scan.

That is, the present invention is a coding method of a quantized DCT block that reduces the number of bits by adaptively changing the scan pattern of each block. The scan pattern is selected so that the line length is as short as possible. Usually, it is determined by using the block of the corresponding motion compensation image that has similar contour characteristics.

The biggest feature of the present invention is that the motion compensation image (Motion Compensated Image) that is common to the encoder and the decoder in the block characteristic analysis of the image so as not to require other additional information, It is compatible with the standard MPEG-2.

When the method of the present invention is applied, a reduction effect of approximately 5% -10% can be expected depending on the bit rate depending on the image.

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

6 and 7 illustrate a flowchart for explaining the method of the present invention, wherein the encoding of the image comprises: initializing a block address to the first block of the image to be encoded (S1); DCT transforming the current block indicated by the block address in the motion compensation image (S2); Calculating a value (t) of the extraction function (F) to grasp the contour characteristic of the DCT block generated in the step (S2) (S3); Determining (I) a scan pattern suitable for the current DCT block from a set of scan patterns defined using the t value calculated in the step S3 and a selection function X; DCT transforming and quantizing the current block of the motion compensated prediction difference image (S5); A variable length coding of the DCT coefficients of the block generated in the step S5 in the order of the scan pattern determined in the step S4, and outputting them as a bit string (S6); Incrementing the address of the block (S7); Detecting whether the end of the effective block is the end of the encoding process, if not the end of the effective block characterized in that it consists of a step (S8) to repeat the process after the step (S2), and the process of decoding the image Initializing a block address to a first block of an image to be decoded (S11); DCT transforming the current block indicated by the block address in the motion compensation image (S12); Calculating a value (t) of the extraction function (F) to grasp the contour characteristic of the DCT block generated in the step (S12) (S13); Determining (I) a suitable scan pattern from a set of scan patterns defined using the t value calculated in the step S13 and a selection function X; Outputting DCT coefficients by variable length coding the compressed bit stream coming from the encoder (S15); Reconstructing the block with the DCT coefficients decoded in step S15 by referring to the scan pattern determined in step S14 (S16); Incrementing the address of the block (S17); Detecting whether or not the end of the effective block is the end of the decoding process, and if not the end of the effective block characterized in that the step (S18) to repeat the process after the step (S12).

8 and 9 show an embodiment of an encoder and a decoder used in the apparatus of the present invention, wherein the encoder 100 includes: a scan pattern storage unit 11 for storing arbitrary scan pattern set data; A DCT converter 12 for receiving a current block of the motion compensation image and performing DCT conversion; A feature extractor (13) which receives the DCT block from the DCT converter (12) and calculates a value of the extraction function (F) to grasp the outline characteristic of the block; A scan pattern determination unit 14 for determining (I) the scan pattern based on the value of the extraction function F output from the feature extraction unit 13; A processor unit 15 applying the quantized DCT block of the motion compensation prediction difference image with reference to the scan pattern determined by the scan pattern storage unit 14 in the scan pattern storage unit 11; And an entropy encoder 16 for variable length encoding the DCT coefficients from the processor 15. The decoder 200 includes a scan pattern storage unit for storing arbitrary scan pattern set data. 21; A DCT converter 22 for receiving a current block of the motion compensation image and performing DCT conversion; A feature extracting unit 23 for calculating an extracting function F value to grasp the outline of the block; A scan pattern determination unit 24 for determining a scan pattern (I) based on an extraction function (F) value output from the feature extraction unit (23); An entropy decoder 25 for variable length decoding the chopped bit stream; The scan pattern determined by the scan pattern determiner 24 is referred to by the scan pattern storage unit 21 is composed of a processor unit 26 for reconstructing the DCT coefficients from the entropy decoder 25 into a block do.

Referring to the effects of the present invention as follows.

First, the basic characteristics of the present invention, the contour image of the original image, the motion compensation image and the motion compensation pre-scheduled difference image is similar to each other. 1, 2, and 3 illustrate examples of the original image, the motion compensation image, and the motion compensation prediction difference image.

In addition, in order to apply the block-adaptive scan method, an arbitrary scan pattern set that reflects the characteristics of the DCT coefficients according to the direction of the contour must be defined.

4 illustrates an example of a set of five scan patterns, but various scan pattern sets may be configured.

The sets of scan patterns are stored in the scan pattern storage units 11 and 21 installed in the encoder 100 and the decoder 200, respectively, and the scan patterns are determined by the scan pattern determination units 14 and 24. If determined (I), the processor unit 15 refers to this.

Meanwhile, in the present invention, in order to determine a suitable scan pattern from a set of scan patterns for each DCT transformed block of the motion compensated prediction difference image, the encoder 100 and the decoder 200 commonly use a valid motion compensation image. In order to make the contour characteristic more efficient, the DCT transform of the block of motion compensation image is used for simple extraction, and then the coefficient is analyzed and used.

That is, the DCT converter 12 or 22 receives the current block indicated by the block address in the motion compensation image and performs DCT conversion.

In addition, the feature extractors 13 and 23 that receive the output signal of the DCT converter 12 and 22 calculate the value t of the extraction function F to determine the contour characteristic of the block. do.

The scan pattern determination unit 14 or 24 determines the scan pattern based on the extraction function F value t and the selection function X output from the feature extraction units 13 and 23 (I). )

5 shows the positions of the DCT coefficients used for the feature extraction in the feature extractors 13 and 23, respectively, where Hi represents the horizontal contour characteristics and Vi represents the positions of the coefficients representing the vertical contour characteristics. If the initial value of the scanning pattern is expressed as I for P, which is a block of the DCT transformed motion compensation image, I is defined as follows.

I (P) = X (F (P))

Here, F is a function for extracting a required characteristic from P, and X is a selection function corresponding to the scan pattern according to the value calculated in F (P).

The function F, which extracts the characteristics of the contour, is the ratio of the states of the horizontal and vertical energies.

In this case, the scan pattern selection function X should be determined to select 1 or 2 for the vertical contour and 4 or 5 for the horizontal contour, as seen in the scan pattern set of FIG. The selection function X for the scan pattern set in FIG. 4 is defined as follows.

In the above equation, ε1 and ε2 are thresholds determined by considering the motion compensation prediction difference image and the extraction function (F).

In addition, the processor unit 16 of the encoder 100 obtains the scan pattern determined by the scan pattern determiner 14 from the scan pattern storage unit 11 and indicates the sage indicated by the block address in the motion compensation prediction difference image. Is then applied to the quantized DCT block of.

As described above, when the DCT blocks quantized by the processor unit 15 are output in the order of the scan patterns, the entropy encoder 16 receives the variable length codings of the DCT coefficients by variable length and exports them into bit strings. After that, the block address is increased, and this exaggeration is repeated until the valid block ends.

On the other hand, the entropy decoder 25 in the decoder 200 decodes a variable length bit stream from the encoder to a variable length to export DCT coefficients.

The processor unit 26 that receives the DCT coefficients from the entropy decoding unit 25 reconstructs the DCT block with reference to the scan pattern selected by the scan pattern storage unit 21 and increases the block address. Repeats until the end of the valid block.

As described above, according to the present invention, an arbitrary set of scan patterns is determined without uniformly applying the scan pattern reflecting the general characteristics of the image as in the prior art, and the image is analyzed for the characteristics of each DCT block and the scan pattern suitable for the same. By applying this, the efficiency can be greatly improved.

Claims (5)

A system in which a method of encoding a predetermined image and decoding a coded image into an original image is applied, the image encoding process includes: initializing a block address with a first block of an image to be encoded (S1); DCT transforming the current block indicated by the block address in the motion compensation image (S2); Calculating a value (t) of the extraction function (F) to grasp the contour characteristic of the DCT block generated in the step (S2) (S3); Determining (I) a suitable scan pattern for the sage DCT block from a set of scan patterns defined using the t value calculated in the step S3 and the selection function X; DCT transforming and quantizing the current block of the motion compensated prediction difference image (S5); (S6) encoding the DCY coefficients of the block generated in step S5 to variable length according to the scan pattern determined in step S4; The block increasing the address (S7); When the end of the valid block is detected and the end of the valid block is completed, the encoding assumption is terminated. If the end of the valid block is not the end of the valid block, step S2 is repeated to perform the process after step S2. Initializing a block address with the first block of S11; DCT transforming the current block indicated by the block address in the motion compensation image (S12); Calculating a value (t) of the extraction function (F) to grasp the contour characteristic of the DCT block generated in the step (S12) (S13); Determining (I) a suitable scan pattern from a set of scan patterns defined using the t value calculated in the step S13 and a selection function X; Reconstructing the block with DCT coefficients by decoding the compressed bit string to a variable length (S16); Incrementing the address of the block (S17); Detecting whether or not the end of the valid block is terminated, and ending the decoding process; if not, the step (S17); Detecting whether or not the end of the valid block is finished, the decoding process is terminated, and if it is not the end of the valid block, step S18 is repeated to the subsequent step (S12). CCT (DCT) block coding method. The method of claim 1, wherein the scan pattern determination I (P) for the block P of the DCT-converted motion compensation image obtains the value of the extraction function F for the block P, and scans the value. A method of coding a DCT block using a block-adaptive scan, characterized in that the pattern selection function (X) value is defined as I (P) = X ((F (P)). The method of claim 1, wherein the extraction function (F) is A coding method of a DCT block using a block-adaptive scan, characterized in that defined as. The method of claim 1, wherein the selection function (X) is determined by threshold values (ε1, ε2) determined in consideration of a motion compensation prediction difference image, an extraction function (F), and the like. A method of encoding a DCT block. A scan pattern storage unit 11 storing arbitrary scan pattern set data, a DCT converter 12 for receiving a DCT transform block of a motion compensation image, and an output signal from the DCT converter 12 Scan pattern based on the value of the extraction function (F) output from the feature extraction unit 13 and the feature extraction unit 13 to calculate the value of the extraction function (F) in order to determine the contour characteristics of the block Scan pattern correction unit 14 for determining (I) and a scan pattern determined by the scan pattern storage unit 11 with the scan pattern determined by the scan pattern determination unit 14 from the scan pattern storage unit 11 An encoder 100 configured to apply a quantized DCT block of the compensated prediction difference image, an entropy encoder 16 to variable length encode DCT coefficients input from the processor 15, and an arbitrary Storing scan pattern set data The scan pattern storage unit 21, the DCT converter 22 which receives the current block of the motion compensation image, and performs DCT conversion, and the feature extractor which calculates an extracting function (F) to grasp the outline of the block ( 23); A scan pattern determiner 24 for determining a scan pattern (I) based on the value of the extraction function F output from the feature extractor 23, and an entropy decoder 25 for variable length decoding the compressed bit stream. Wow; The processor unit 26 reconstructs the DCT coefficients decoded through the entropy decoding unit 25 into the block pattern according to the selected scan pattern by referring to the scan pattern determined by the scan pattern determiner 24 by the scan pattern storage unit 21. (DCT) block encoding apparatus using a block-adaptive scan, characterized in that consisting of).