KR0130167B1 - Mpeg apparatus - Google Patents

Abstract

The apparatus for increasing compression ratio of image without degrading image quality includes a compensator(21) performing quantization and discrete cosine transformation(DCT) of input image to compensate motion, a coding circuit(22) coding output of the compensator with variable length coding method, an circuit(23) providing motion vector by estimating motion of imput image, a converter(31) aligning the motion vector to 2 dimension of image, a circuit(32) coding output of the converter without loss, a circuit(33) coding output of the converter as considering the coding loss, and a selector(34) selecting one of output of the circuits(32, 33) with control signal and providing to multiplexor/output buffer(25).

Description

Video compression encoding device

1 is a block diagram of a conventional video compression encoding apparatus.

2 is a block diagram of a video compression encoding apparatus of the present invention.

3 is a block diagram of a motion vector compression unit for realizing the lossless coding of the present invention.

4 is a block diagram of a motion vector compression unit for realizing lossy coding of the present invention.

5 is a block diagram of a motion V vector compression unit for selectively using lossless coding and lossy coding of the present invention.

6 is a diagram in which the video compression encoding apparatus of the present invention is applied to an MPEG system.

* Explanation of symbols for the main parts of the drawings

21: DCT / quantization / motion compensation unit 22: variable length coding unit

23: motion estimation unit 24: motion vector compression unit

25: multiplexer / output buffer section 31: motion vector compression section

32: lossless coding unit 33: loss sensitive coding unit

34: selection

The present invention relates to a video compression encoding apparatus for transforming and coding a motion vector used to reduce temporal redundancy in video encoding so as to increase the compression ratio of an image without compromising image quality.

In general, video compression encoding provides a high image quality while providing a high compression ratio. In order to obtain a high compression ratio, it is necessary to remove an image in time and space.

P-frames and B-frames are used to reduce temporal redundancy, and motion estimation and motion compensation techniques are used, and the motion vectors are compressed using variable length codes.

In order to reduce the amount of space in the space, the image is divided into block units, and each block is subjected to Discrete Cosine Transform (DCT), which is compressed by quantization and run-length coding.

FIG. 1 shows a conventional video compression encoding apparatus.

According to this conventional technique, the input image is stored in the field / frame memory 1 in field frame units, and the difference from the image immediately before the motion compensation finished in the adaptive predictor 2 is finished. The difference image is obtained from the subtractor 3, and the difference image is DCT and quantized in the DCT unit 4 and the quantization unit 5, and the result is run-length coding and variable length coding in the variable length coding / multiplexer 6. Compressed in a manner and output through the buffer (7).

The output signal of the quantization unit 5 is restored to the original signal through the inverse quantization unit 8 and the inverse DCT unit 9, and the arithmetic unit 10, the field / frame memory 11, and the adaptive predicate ( After the motion is compensated through 2), the image immediately before the motion compensation is completed is input to the subtractor 3 to continuously repeat the video compression, and the motion estimation unit 12 and 13 estimate the motion of the image. The obtained motion vector is coded by the variable length coding method in the variable length coding / multiplexer 6 and then output through the buffer 7.

However, according to the conventional technology as described above, by compressing only the variable length coding for the motion vector, the motion vector in the actual image is separated into the background and the object so that the movement of the object with respect to the background or the motion of the background with respect to the object That is, the shape indicates that the motion vector is closely related to the two-dimensional plane of the image, which indicates that there is much room for compression of the motion vector. In the related art, only the variable length coding is performed on the motion vector, so that the compression rate is low. there was.

The present invention has been devised to solve the above-mentioned conventional problems, and represents a motion vector estimating the motion of an input image as a two-dimensional image, and outputs it by coding it using an intra coding method of lossless coding or loss sensitive coding. Accordingly, an object of the present invention is to provide a motion picture compression encoding apparatus capable of compressing an image at a high compression rate without compromising image quality, which will be described in detail with reference to FIGS. 2 to 6.

2 is a circuit diagram illustrating a video compression encoding apparatus of the present invention, as shown therein, a DCT / quantization / motion compensation unit 21 for DCT and quantization of an input image, and the DCT / quantization / motion compensation unit 21. Variable length coding unit 22 for coding the output signal of the variable length coding method, a motion estimation unit 23 for estimating the motion of the input image and outputting a motion vector, and an output signal of the motion estimating unit 23; Is a motion vector compression unit 24 for compressing the signal by the intra compression method, and a multiplexer / output buffer unit 25 for selectively outputting the output signal of the variable length coding unit 22 or the motion vector compression unit 24. ).

In the video compression encoding apparatus of the present invention configured as described above, the input image is inputted to the DCT / quantization / motion compensation unit 21, DCT and quantized, and compressed into variable length coding by the variable length coding unit 22, and then multiplexer / It is output through the output buffer part 25. FIG.

In addition, the motion of the input image is estimated by the motion estimation unit 23, and a motion vector is output. The output motion vector is input to the motion vector compression unit 24, compressed by an intra compression method, and then multiplexer / output buffer. It is output through the unit 25.

The intra compression method includes lossless coding for lossless coding of a motion vector image, lossy coding method for coding a motion vector image in consideration of loss, and a selector to selectively use the lossless coding method and lossy coding method. There is a way.

The three methods will be described with reference to FIGS. 3 to 5 as follows.

FIG. 3 is a detailed view showing the motion vector compression unit 24 of FIG. 2, which is a configuration diagram to realize lossless coding.

Referring to FIG. 3, the configuration of the motion vector compression unit 24 for realizing lossless coding is as follows.

The motion vector image converter 31 arranging the motion vectors output from the motion estimator 23 into two-dimensional images and outputting horizontal and vertical motion vector images to the lossless coding unit 32, and the input horizontal and vertical motions. It is composed of a lossless coding unit 32 outputting the vector image to the multiplexer / output buffer unit 25 by DPCM coding and variable length coding.

The lossless coding unit 32 is a DPCM execution unit 32-A for obtaining a difference between the vector values of the horizontal and vertical motion vector images input by the DPCM coding method, and the DPCM execution unit 32-A. It consists of a variable length coding section 32-B for variable length coding the output of < RTI ID = 0.0 >

According to the present invention configured as described above, the number of motion vectors output from the motion estimator 23 and input to the motion vector image converter 31 is determined by the block size of the motion estimation. Assuming that the size of the input image inputted by) is 2n × 2n and that the size of the motion estimation block is 16 × 16, the motion vector image converter 31 has 2 (n-4) × 2 for one image. As many as (n-4) motion vector cells are generated.

Here, since the motion vector cell has both vertical and horizontal components, the total number of motion vectors is 2 (n-4) × 2 (n-4) × 2, and the motion vector image converter 31 converts the motion vectors. A motion vector image having two 2 (n-4) × 2 (n-4) sizes, that is, a two-dimensional image, is output.

At this time, the DPCM performing unit 32-A of the lossless coding unit 32 regards the horizontal and vertical motion vector image input from the motion vector image converting unit 31 as a normal image and uses each vector as a conventional DPCM method. The difference between the values is obtained, and the variable length coding unit 32-B performs variable length coding on the difference value and outputs the difference value to the multiplex / output buffer unit 25.

When decoding a lossless coded image as described above, the motion vector obtained by decoding the lossless coded motion vector compressed bit stream by a uniform method can be reproduced and used as it is.

FIG. 4 is a detailed view showing the motion vector compression unit 24 of FIG. 2, which is a configuration for realizing loss sensitive coding.

Referring to FIG. 4, the lossy loss coding unit 33 of the motion vector compression unit 24 for realizing lossy loss coding will be described in detail. The 2D image of the motion vector output from the motion estimation unit 23 will be described. A motion vector image converter 31 for outputting horizontal and vertical motion vector images to the DCT / quantizer 33-A and a horizontal and vertical motion vector image input from the motion vector image converter 31. DCT and quantize the output length of the DCT / quantizer 33-A and the output signal of the DCT / quantizer 33-A to quantize the variable length coding unit 33-C, and multiplexer / output buffer. Inverse DCT and inverse quantization of the variable-length coding unit 33-C and the DCT / quantization unit 33-A outputted to the unit 25 and reconstructing the motion vector image to DCT / quantization / Inverse DCT / quantization unit 33-B output to the motion compensation unit 21.

The DCT / quantizer 33-A is a motion vector image DCT unit 33-A1 for DCT processing the input horizontal and vertical motion vector images, and a motion vector image quantizer 33-A2 for quantization. And the inverse DCT / quantizer 33-B includes a motion vector image inverse quantizer 33-B2 for inversely quantizing the motion vector image coded by the DCT / quantizer 33-A. It consists of a motion vector video region DCT section 33-B1 to be processed.

In the present invention configured as described above, the loss sensitive coding unit 33 uses loss sensitive coding. In this case, the horizontal and vertical motion vector images inputted to the input have a value between -N and N according to the size of the motion search region (N). In addition, this can be regarded as a normal image having a dynamic range of -N to N, and is DCT and quantized through the motion vector image DCT unit 33-A1 and the motion vector image quantization unit 33-A2. The variable length coding unit 33-C is variable length coded and output to the multiplexer / output buffer unit 25.

In addition, the coded motion vector image is reconstructed as a motion vector image by the motion vector image inverse quantizer 33-B2 and the motion vector image inverse DCT unit 33-B1, wherein the reconstructed motion vector is the first motion. Since the vector image is not the same by loss, the information is fed back to the DCT / quantization / motion compensation unit 21, and motion coding, DCT, and quantization are performed on the basis of the entire coding.

FIG. 5 illustrates a method of selectively using lossless coding and lossy coding, and the configuration of the lossless coding unit 32 of FIG. 3 and the lossy coding unit 33 of FIG. It comprises a selection unit 34 which selects one output value as a control signal and outputs the selected value to the multiplexer / output buffer unit 25.

According to the present invention configured as described above, the selector 34 programs values output from the variable length coding units 32-B and 33-C of the lossless coding unit 32 and the loss sensitive coding unit 33. One of the two input values is selected as the control signal set in advance, and the selected value is output to the multiplexer / output buffer unit 25.

The present invention described above should be compatible with the existing MPEG system. As shown in FIG. 6, the compressed bit string output from the MPEG encoder 41 is input to the compression unit 42 of the present invention. Finds a motion vector from the input compressed bit string, generates and stores the compressed and new compressed bit strings as described above, and the received compressed bit string is decompressed by the decompression unit 43 of the present invention and the MPEG decoder 44 Decode from

As described above, the present invention converts a motion vector into a two-dimensional image and codes the image by an intra coding method so that the image can be compressed at a high compression rate without compromising image quality. Better image quality can be obtained.

Claims (4)

DCT / quantization / motion compensation unit 21 for performing motion compensation by DCT and quantizing the input image, and variable length coding unit for coding the output signal of the DCT / quantization / motion compensation unit 21 with variable length coding method 22, a motion estimator 23 for estimating the motion of the input image and outputting a motion vector, and a motion vector image converter for arranging the motion vectors output from the motion estimator 23 into a two-dimensional image ( 31), the lossless coding unit 32 for coding the output signal of the motion vector image conversion unit 31 without loss, and the loss occurring in coding the output signal of the motion vector image conversion unit 31 in consideration of The loss sensitive coding unit 33 to be coded and the output signal of the lossless coding unit 32 or the loss sensitive coding unit 33 are selected according to a preset control signal and outputted to the multiplexer / output buffer unit 25. A selection unit 34 and the selection unit ( And a multiplexer / output buffer unit (25) for selectively outputting the lossy-sensitive coding or lossless-considered coding signal selected at (34) or the output of the variable length coding unit (22). 2. The DCT / quantizer 33-A of claim 1, wherein the loss sensitive coding unit 33 is a DCT / quantizer 33-A for outputting DCT and quantized horizontal and vertical motion vector images input from the motion vector image converter 31. The variable length coding unit 33-C outputs the variable length coding of the output signal of the DCT / quantization unit 33-A to the multiplexer / output buffer unit 25, and the DCT / quantization unit 33-A. Video compression encoding, comprising: a DCT / quantization unit 33-B for inverse DCT and inverse quantization coded in the N-D, and restoring the motion vector projection to a DCT / quantization motion compensator 21 Device The DPCM execution unit 32 of claim 1, wherein the lossless coding unit 32 calculates a difference between the vector values of the horizontal and vertical motion vector images input from the motion vector image conversion unit 31 by DPCM coding. -A) and a variable length coding unit 32-B for variable length coding the output from the DPCM performing unit 32-A and outputting the variable length coding unit to the multiplexer / output buffer unit 25. Encoding device 3. The DCT / quantization unit 33-A according to claim 2, wherein the DCT / quantization unit 33-A includes a motion vector image DCT unit 33-A1 for DCT processing the input horizontal and vertical motion vector images, and the motion vector image DCT unit 33-A. A1) comprises a motion vector image quantizer 33-A2 for quantizing the DCT-processed horizontal and vertical motion vector image, wherein the inverse DCT / quantizer 33-B is the DCT / quantizer 33-A. And a moving vector image inverse quantization unit (33-B2) for inverse quantization of the motion vector image coded in (D)) and a motion vector image inverse DCT unit (33-B1) for inverse DCT processing.