KR20050052130A - Image decoding method - Google Patents

Abstract

The present invention relates to a method for decoding an image, the method comprising: a first step of determining whether a macroblock to be decoded of a slice including a plurality of macroblocks is a P macroblock; If the macroblock to be decoded is a P macroblock, finding a motion vector corresponding to the macroblock to be decoded in the reference frame; A third step of determining whether the macroblock having found the motion vector in the second step is the last macroblock in the slice; After the third step, if the macroblock continuously finding the motion vector in the third step is not the last macroblock in the slice, and the macroblock to be decoded in the first step is a P macroblock, the first to third steps A fourth step of continuing to perform; If the macroblock performing the third step in the fourth step is the last macroblock in the slice, among the macroblocks performing the second step, macroblocks having the same motion vector are merged and interpolated. ) And a fifth step of decoding.

Accordingly, the present invention can prevent data from being read redundantly by contiguous, having the same motion vector, and the reference frame merging the same blocks to perform motion compensation, thereby reducing the use of memory and reducing power consumption. It can be effective.

Description

Image decoding method

The present invention relates to a method for decoding an image, and more particularly, by concatenating adjacent, having the same motion vector, and performing reference to the same frame by combining the same blocks, the data can be prevented from being read repeatedly. The present invention relates to a method of decoding an image, which can reduce the use of memory and reduce power consumption.

Currently, in the video compression method, in order to code a macro block, the compression rate is found by finding a part similar to the current macro block in the previous frame that has been coded, and transmitting the position, and coding the difference between the part and the current macro block. Is raising.

In MPEG (Moving Picture Experts Group) 2 video coding, motion vectors are processed in units of 16x16 macroblocks, while MPEG4 Part2 video coding allows the use of motion vectors in blocks of 8x8.

Furthermore, in the case of MPEG4 Part 10, a method of dividing a macro block into a maximum of 16 4x4 blocks and processing them using motion vectors is adopted.

The smaller the size of the block to which such a motion vector is allocated is, the smaller the size of the difference data becomes and the compression efficiency becomes higher.

However, as the size of a block decreases, the number of motion vectors may increase, and as the size of a visiting block decreases, the number of memory accesses increases.

1 is a schematic block diagram of a typical video decoder for MPEG, which is a variable length decoder (VLD) for variable length decoding a compressed bit stream into motion vectors, quantization values, and discrete cosine transform (DCT) coefficients. Wow; An inverse quantization unit 20 for inverse quantization of the DCT coefficients separated from the VLD; An inverse discrete cosine transform of the inverse quantized DCT coefficients by the inverse quantization unit 20; A motion compensator (50) for motion compensation using the motion vectors extracted from the VLD (10); The IDCT 30 includes an adder 40 for outputting a decoded image by adding the inverse discrete cosine transformed data and the compensated data to the motion compensator 50.

The video decoder for MPEG is configured to decode the compressed bit stream variable length in the VLD (10) to separate the motion vector, the quantization value and the DCT coefficients, and inverse quantizer 20 separates the DCT coefficients separated from the VLD (10). Inverse quantization, and inverse discrete cosine transform the DCT coefficient dequantized by the dequantizer 20 in the IDCT (30).

The motion compensation unit 50 performs motion compensation using the motion vector extracted from the VLD 10, and the data compensated in the motion compensation unit 50 and the inverse discrete cosine transform in the IDCT 30. The added data is added by the adder 40 and output as a decoded image.

FIG. 2 is a view schematically illustrating a data configuration method for reconstructing blocks of an image according to the prior art. In general, the size of data required for reconstructing one block constitutes the size of the block and data required for reconstruction. It depends on the method.

When the position referenced by the current decoding block 61 is an integer position, data corresponding to the size of the current block ('a' in FIG. 2) is required.

When the position of the block 62 to be decoded is referred to as a half pixel or a quarter pixel instead of an integer, a larger amount of data (corresponding to 'b' in FIG. 2) may be obtained. You should note it and filter it out.

In addition, if two adjacent blocks 63 and 64 have the same motion vector and reference the same reference picture, in the conventional method, each of the individual blocks 63 and 64 reads data from memory. Process.

At this time, the data read from the memory by each block, such as the hatched portion of 'c' of FIG.

Therefore, when the same motion vector is referenced and the same reference image is referred to, an unnecessary reading amount of data is increased, resulting in an increase in the use of memory and an increase in power consumption.

Accordingly, the present invention has been devised to solve the above-described problems. The present invention can prevent data from being read redundantly by performing motion compensation by concatenating adjacent blocks with the same motion vector and the same reference frame. Therefore, the object of the present invention is to provide a method of decoding an image which can reduce the use of memory and reduce power consumption.

A preferred aspect for achieving the above object of the present invention comprises: a first step of determining whether a macroblock to be decoded of a slice comprised of a plurality of macroblocks is a P macroblock;

If the macroblock to be decoded is a P macroblock, finding a motion vector corresponding to the macroblock to be decoded in the reference frame;

A third step of determining whether the macroblock having found the motion vector in the second step is the last macroblock in the slice;

After the third step, if the macroblock continuously finding the motion vector in the third step is not the last macroblock in the slice, and the macroblock to be decoded in the first step is a P macroblock, the first to third steps A fourth step of continuing to perform;

If the macroblock performing the third step in the fourth step is the last macroblock in the slice, among the macroblocks performing the second step, macroblocks having the same motion vector are merged and interpolated. A method of decoding an image comprising a fifth step of decoding) is provided.

Another preferred aspect for achieving the above object of the present invention includes a first step of determining whether a macroblock to be decoded of a slice composed of a plurality of macroblocks is included in a picture and is a P macroblock; ;

If the macroblock to be decoded is a P macroblock, finding a motion vector corresponding to the macroblock to be decoded in the reference frame;

A third step of determining whether the macroblock having found the motion vector in the second step is the last macroblock in the slice;

 After the third step, if the macroblock continuously finding the motion vector in the third step is not the last macroblock in the slice, and the macroblock to be decoded in the first step is a P macroblock, the first to third steps A fourth step of continuing to perform;

If the macroblock performing the first step in the fourth step is an I macroblock instead of a P macroblock, among the macroblocks performing the second step, macroblocks having the same motion vector are merged. A fifth step of interpolating and decoding;

In the fifth step, a method for decoding an image comprising a sixth step of decoding the I macroblock by referring to data in the decoded macroblock is provided.

Hereinafter, exemplary embodiments of the present invention will be described with reference to the accompanying drawings.

3 is a flowchart illustrating a method of decoding an image according to the present invention. Before describing FIG. 3, one picture is composed of a plurality of slices, and the slice is a plurality of slices. It is composed of P macroblocks or a plurality of P macroblocks and I macroblocks.

Here, the P macroblock is a macroblock referencing data in the previous frame, and the I macroblock is a macroblock referencing data in the currently decoded macroblock.

3 of the present invention is a method of decoding one slice included in a picture. First, it is determined whether a macroblock to be decoded for a slice included in the picture and composed of a plurality of macroblocks is a P macroblock. S100 step)

Thus, performing step S100, as shown in Figures 4a and 4b, the macro block to be decoded in the slice (200, 300), as shown in Figure 4a, the P macroblock 201 and as shown in Figure 4b, as shown in Figure 4b This is because there is a case of block 304.

Thereafter, if the macroblock to be decoded is a P macroblock, a motion vector corresponding to the macroblock to be decoded in the reference frame is found (step S110).

Subsequently, it is determined whether the macroblock having found the motion vector is the last macroblock in the slice by performing step S110 (step S120).

Subsequently, in step S120, if the macroblock that finds the motion vector is not the last macroblock in the slice, and the macroblock to be decoded in step S100 is a P macroblock, that is, as shown in FIG. 4A, macroblocks of the slice are continuously In the case of the P macroblock, steps S100 to S120 are continued.

Next, if the macroblock performing step S110 is the last macroblock 216 as shown in FIG. 4A, among the macroblocks performing step S110 before, macroblocks having the same motion vector are merged. And interpolation to decode (step S130).

In this case, the merging may be performed by adjacent blocks and reference frames further satisfying the same condition.

On the other hand, in step S100, if the macroblock to be decoded is not a P macroblock, as shown in FIG. 4B, since the I macroblock 304 is used, macroblocks previously found a motion vector (that is, P macroblocks). ) Are merged, interpolated and decoded (step S140).

Here, the macroblocks 301, 302, and 303 before the I macroblock 304 are continuous P macroblocks, and thus are macroblocks obtained by continuously performing steps S100 to S120 to find a motion vector.

Therefore, when described in detail with reference to FIG. 4B, in step S140, merging and interpolation include one macroblock 301 whose motion vectors are not identical among the previous macroblocks 301, 302, 303 of the I macroblock 304. The interpolation is performed by merging and interpolating macroblocks 302 and 303 having the same motion vector.

Thereafter, the I macroblock 304 is decoded by referring to the data in the currently decoded macroblocks 301, 302 and 303 (step S150).

Then, it is determined whether the decoded I macroblock is the last macroblock in the slice (step S160), and if the last macroblock is terminated, otherwise, the step S100 is performed.

5 is a view illustrating a merged shape of macroblocks in a slice according to the present invention, and there are a plurality of macroblocks in the slice 400.

When decoding these macroblocks by performing the method of the present invention, as shown in Fig. 5, the macroblock to be decoded is the I macroblock 411, and the macroblocks having the same motion vector are '406', '407'. And '410', only the rectangular '406' and '410' macroblocks are merged to interpolate.

That is, in the present invention, the shape of the merged macroblock is preferably rectangular.

As described above, the present invention can prevent data from being redundantly read by contiguous, having the same motion vector, and the reference frame merging the same blocks to perform motion compensation, thereby reducing the use of memory, There is an effect that can reduce the power consumption.

Although the invention has been described in detail only with respect to specific examples, it will be apparent to those skilled in the art that various modifications and variations are possible within the spirit of the invention, and such modifications and variations belong to the appended claims.

1 is a schematic block diagram of a typical video decoder for MPEG

2 is a view schematically showing a data configuration method for reconstructing blocks of an image according to the prior art;

3 is a flowchart illustrating a method of decoding an image according to the present invention.

4A and 4B illustrate macroblocks in a slice in accordance with the present invention.

FIG. 5 illustrates a shape in which macroblocks in a slice are merged according to the present invention. FIG.

<Description of the symbols for the main parts of the drawings>

10: VLD (Variable Length Decoder) 20: Inverse quantization part

30: Inverse Discrete Cosine Transform (IDCT)

40: adder 50: motion compensation unit

200,300,400: Slice 201: P macroblock

216,301,302,303,406,407,410: Macroblock

304,411: I macroblocks

Claims (4)

A first step of determining whether a macroblock to be decoded of a slice, which is included in a picture, is a P macroblock; If the macroblock to be decoded is a P macroblock, finding a motion vector corresponding to the macroblock to be decoded in the reference frame; A third step of determining whether the macroblock having found the motion vector in the second step is the last macroblock in the slice;  After the third step, if the macroblock continuously finding the motion vector in the third step is not the last macroblock in the slice, and the macroblock to be decoded in the first step is a P macroblock, the first to third steps A fourth step of continuing to perform; If the macroblock performing the third step in the fourth step is the last macroblock in the slice, among the macroblocks performing the second step, macroblocks having the same motion vector are merged and interpolated. And a fifth step of decoding the video. A first step of determining whether a macroblock to be decoded of a slice, which is included in a picture, is a P macroblock; If the macroblock to be decoded is a P macroblock, finding a motion vector corresponding to the macroblock to be decoded in the reference frame; A third step of determining whether the macroblock having found the motion vector in the second step is the last macroblock in the slice;  After the third step, if the macroblock continuously finding the motion vector in the third step is not the last macroblock in the slice, and the macroblock to be decoded in the first step is a P macroblock, the first to third steps may be performed. A fourth step of continuing; If the macroblock performing the first step in the fourth step is an I macroblock instead of a P macroblock, among the macroblocks performing the second step, macroblocks having the same motion vector are merged. A fifth step of interpolating and decoding; And a sixth step of decoding the I macroblock by referring to data in the decoded macroblock. The method according to claim 1 or 2, The merging is performed by a block adjacent to each other, and the reference frame is further satisfied with the same condition. The method according to claim 1 or 2, And the shape of the merged macroblock is rectangular.