KR100640937B1 - Method of image compression for video decoding based on motion compensation - Google Patents

Abstract

The present invention relates to an image processing method, and more particularly, to an image compression algorithm for motion compensation-based image decoding such as MPEG2.

A method of processing a video image, the method comprising the steps of: determining image characteristics for a predetermined area comprising four consecutive pels in the video image; Encoding or decoding the video image based on the determined image characteristic, wherein determining the image characteristic comprises sequentially comparing one or more calculated values with a reference value, And determining the image characteristic, wherein the one or more calculated values are determined based on a difference value using the pels in the predetermined region.

Therefore, according to the present invention, it is possible to reduce the memory capacity required when performing motion compensation-based video decoding such as MPEG2, to simplify or facilitate memory access, and to reduce the complexity of calculation for encoding or decoding.

Video Compression Method, MPEG2 Video Compression Encoding

Description

Image Compression Method for Motion Compensation Based Image Decoding {METHOD OF IMAGE COMPRESSION FOR VIDEO DECODING BASED ON MOTION COMPENSATION}

1 is a view showing an example of a bundle type for explaining the image compression method of the present invention

2 is a diagram showing an example of a code granting method for explaining an image compression method according to the present invention;

3 illustrates an example of a quantization method for explaining an image compression method according to the present invention.

4 is a diagram illustrating an example of a coding table for explaining an image compression method according to the present invention.

The present invention relates to an image compression method, and more particularly, to an image compression algorithm for motion compensation-based image decoding such as MPEG2; In applying compression coding for each region according to the characteristics of the image, the size of the pel block is defined, the image characteristics are first distinguished from the difference between adjacent pels, and the image characteristics are used to determine the image characteristics. The present invention relates to an image compression method of performing image compression by finally classifying the image type into image types and assigning an appropriate codeword to each image type.

Conventionally, as MPEG2, a video compression method for motion compensation-based video decoding, which compresses an image by applying different coding schemes according to characteristics of regions in an image, thereby efficiently using memory, increasing transmission efficiency, and minimizing image quality. Doing.

In such a video compression method, the simplification of calculation and the efficiency and ease of data access are key.

Conventional image compression methods have been proposed such as tree structure vector quantization (TSVQ), pyramid vector quantization (PSQ), variable length coding (VLC), etc. The efficiency or ease of use is not evenly satisfied, and in most cases, if the efficiency or ease of data access is satisfied, the calculation is complicated, or the calculation is simple, but the efficiency or ease of data access is not satisfied.

The present invention compresses and encodes an image, and calculates a difference between two adjacent pels to divide an image region into a mild region, an edge region, and a random pattern region. A method of compressing and encoding an image by a method of giving a word; Image characteristics are determined primarily from the difference between adjacent pels, and based on this information, each bundle type is determined while finally determining a bundle type in which an image is defined as a characteristic for a predetermined unit area. An image compression method for compressing and encoding an image by a method of appropriately assigning a codeword is provided.

In particular, the present invention sets a size of a pel block to define a bundle, retrieves values between adjacent or neighboring pels and their changing patterns to determine image characteristics, and to determine each image characteristic. Memory compression and memory access required when performing motion compensation-based image decoding such as MPEG2 by compressing and encoding an image in a manner of assigning a codeword for a corresponding region through an appropriate bit allocation. An image compression method is provided to increase the simplicity or ease of use of the algorithm, and to reduce the complexity of calculation for encoding or decoding.

In addition, the present invention defines a bundle by setting the size of the pel block, provides a CDC (Classified Differential Coding) algorithm for retrieving the values between the neighboring or neighboring pels and the pattern of change, the calculation that is a requirement for image compression coding An image compression method is provided to satisfy the complexity and random access performance of a memory.

A method of processing a video image, the method comprising the steps of: determining image characteristics for a predetermined area comprising four consecutive pels in the video image; Encoding or decoding the video image based on the determined image characteristic, and determining the image characteristic comprises sequentially comparing one or more calculated values and a reference value, And determining the image characteristic, wherein the one or more calculated values are determined based on a difference value using the pels in the predetermined region.

Preferably, the comparing step includes a first comparison process based on a first calculated value determined based on a difference between the first pel and the second pel in the predetermined region, and the second pel and the third pel in the predetermined region. A second comparison process based on the second calculated value determined based on the difference of and a third comparison process based on the third calculated value determined based on the difference between the third and fourth pels in the predetermined region. Characterized in that made.

In the determining of the image characteristic, the predetermined region may be viewed as one bundle, and the image characteristic of the predetermined region may be represented as a bundle type.

Hereinafter, with reference to the accompanying drawings, preferred embodiments of the present invention that can specifically realize the above object will be described. The same components as in the prior art are given the same names and the same reference numerals for convenience of description, and detailed description thereof will be omitted.

The video image processing method according to the present invention is characterized by comprising a process of determining an image bundle type and assigning a codeword to each of the determined image bundle types through appropriate bit allocation.

Preferably, the process of determining the image bundle type may classify an image into a mild region, an edge region, and a random pattern region based on human visual characteristics (elements affecting image recognition). The image compression is performed by finally determining the characteristics of the image from the difference and the change pattern between adjacent or neighboring pels for the region, and assigning an appropriate codeword corresponding to the pattern of each pel change through bit allocation. It features.

[One]. First, in the present invention, a bundle is defined.

The bundle defines width x length as 4pels x 1pel.

If the size of the pel block is too large, it is difficult to find a feature. If the size of the pel block is too small, the coding gain is reduced, so that 4 × 1 is defined as the size of the bundle.

[2]. The difference between adjacent pels in the defined pel block (1 bundle) is determined first to determine whether the bundle type belongs to a mild, an edge, or a random pattern.

For example, if the difference between adjacent pels is greater than the reference value Th and the reference value is exceeded, the value change between adjacent pels may be large, so it may be edge or random, otherwise it may be viewed as mild.

From this difference between neighboring (or neighboring) pels, we defined the jump indicator J _ij as

Since 1 bundle is defined as 4 pels (p ₀ , p ₁ , p ₂ , p ₃ ; p _i , 0 ≦ _i ≦ ₃ ), the adjacent pels (p ₀ -p ₁ , p ₁ -p ₂ , p ₂ − p ₃ ) is compared with a predetermined preset reference value J _T ,

(a). If p _j -p _i ＞ J _T, J _ij = +

(b). If absolute value (p _j -p _i ) ≤ J _T, then T _ij = 0 (there is little change in Pel value)

(c). If p _j -p _i <J _T, then T _ij =-(the change in the Pel value decreases more than the predetermined standard value)

Therefore, obtain the jump indicators (J ₀₁ , J ₁₂ , J ₂₃ ) for the preceding adjacent pels (p ₀ -p ₁ , p ₁ -p ₂ , p ₂ -p ₃ ) and determine which value is the value. We can classify the bundle types into roughly mild bundles, edge bundles, and random bundles, and more specifically, mild, random bundles, and edge bundles, each of which can be classified into 12 cases. Eight code types Tc shown in FIG. 1 are assigned to each bundle type.

This process is also performed for the neighboring pels in the bundle to finally determine how the bundle type and how much the pel value changes in the bundle, ie the final characteristics of the image.

[3]. As described above, the adjacent pels obtained jump indicators from the difference between neighboring pels, classified bundle types from them, and determined eight code types for each case corresponding to each bundle type.

A codeword suitable for each characteristic is appropriately selected and used for the code type determined as described above. In the present invention, an anchor code (Ca), a prediction code (Cp), and a slope code (Cs) are used. Compression coding is completed by selecting a code suitable for an image characteristic from a slope compensation code (Cc) and performing appropriate bit allocation to each code.

2 shows each of these codes.

In the case of (a) of FIG. 2, the difference between the predicted pel and the predicted pel as an anchor pel is set as a prediction code (Cp), and the predicted from the anchor pel as shown in (b). If the interpolated pel is within the predetermined range between the pels, the slope between the anchor and the predicted pel is given as the code value (Cs), and as shown in (c), the pel to be compensated and restored is If it does not fall within the predetermined slope range, the difference between the two is given by a slope compensation code (Cc).

[4]. In this way the code type (Tc) as shown in Figure 1 has a change in value between adjacent pel naejineun peldeul neighboring (changes the degree and direction), one end were divided into 8 types, anchor pel (p _a) and the compensation pel where (p _i ), the slope pel (p _s ), and the code type (Tc) three additional search for the following cases to give three additional code type (Tc).

That is, when Cs = p _s -p _a , q _i = p _a + Cs / 2,

If the absolute value (p _i -q _i )> E _{i , max at} Tc = 3, change the code type to Tc = 11,

If the absolute value (p _i -q _i )> E _{i , max} at Tc = 4, change the code type to Tc = 12.

If the absolute value (p _i -q _i )> E _{i , max at} Tc = 7 then change the code type to Tc = 14.

Where q _i is the restored value of the compensation pel p _i and E _{i , max} represents the maximum allowable error of p _i .

In FIG. 3A, the quantization range Qp mapped to the quantization range for the prediction code is shown, and in FIG. 3B, the quantization error and quantization value Qc for the code according to the slope compensation are shown.

In Fig. 3A, the quantization error Ep is in the range of -4 &lt; Ep &lt; 4, and in Fig. 3B, the quantization error Ec is in the range of -4 &lt; Ep &lt;

[5]. Thus, in the present invention, a total of 11 code types have been determined. Compression coding is completed by performing bit allocation as shown in FIG.

In other words, if the code type Tc = 0, the bundle type is mild. In this case, 4 bits for the code type (Tc), 8 bits for the anchor code (Ca), and 4 bits for each of the three prediction codes (Cp). Assigned to each.

If the code type Tc = 1, the bundle type is an edge. In this case, 4 bits for the code type Tc, 8 bits for the anchor code Ca, and 4 bits for the two prediction codes Cp, respectively. 6 bits are allocated by the anchor code Ca.

In the same way, the code type (Tc), anchor code (Ca), prediction code (Cp), slope code (Cs), and slope compensation code (Cc) are appropriately assigned to the remaining code types, respectively. Image compression data was obtained.

According to the video image processing method according to the present invention, by performing image compression encoding using the CDC algorithm, the complexity of calculation required for image data compression can be reduced, and the ease of random access of the memory can be ensured.

Claims (8)

In the method for processing a video image, Determining image characteristics for a given area comprising four consecutive pels in the video image; Encoding or decoding the video image based on the determined image characteristic; Determining the image characteristics, And comparing the at least one calculated value with a reference value sequentially, and determining the image characteristic according to the comparison result, wherein the at least one calculated value is determined based on a difference value using the pels in the predetermined region. Video image processing method. The method of claim 1, wherein the comparing step, A first comparison process based on a first calculated value determined based on the difference between the first pel and the second pel in the predetermined region, A second comparison process based on a second calculated value determined based on the difference between the second pel and the third pel in the predetermined region, And a third comparison process based on a third calculation value determined based on the difference between the third pel and the fourth pel in the predetermined region. The method of claim 2, wherein determining the image characteristic comprises: The video image processing method according to claim 1, wherein the predetermined area is viewed as one bundle, and the image characteristic of the predetermined area is represented as a bundle type. The method of claim 3, wherein The bundle type, And allocating bits in a combination using at least one of a bundle type code, an anchor code, a prediction code, a gradient code, and a compensation code. The method of claim 4, wherein the bundle type, And a mild bundle type, a random pattern bundle type, and an edge bundle type. The method of claim 5, Mild bundles and random bundles each map to one code type, The edge bundles are mapped by assigning bits to at least one combination of an anchor code, a prediction code, a slope code, and a slope compensation code, respectively, corresponding to changes in neighboring or neighboring pel values and the change direction thereof. Video image processing method. The method of claim 1, Comparing the at least one calculated value and the reference value sequentially, Comparing the first calculated value determined based on the difference between the first pel and the second pel and a predetermined reference value, the comparison of the second calculated value determined based on the difference between the second pel and the third pel and the predetermined reference value, and And a comparison between the third calculated value determined based on the difference between the third pel and the fourth pel and a predetermined reference value. The method of claim 1, wherein the one or more calculated values are: Video image processing characterized in that it is determined based on the difference between the first and second pels, the difference between the second and third pels, and the difference between the third and fourth pels in the predetermined region. Way.

Images