KR100232182B1 - Apparatus and method for image compression of digital tv - Google Patents

Abstract

The present invention relates to a video compression apparatus and a method of a directive using a motion compensated DCT coding, and a variable length decoder for decoding and outputting DC coefficients and motion vectors in a compressed and transmitted bitstream, and outputting the variable length decoder. The DC error block output unit outputs the resultant DC filtering by performing coarse filtering, dequantizes the remaining coefficients, and performs an inverse discrete cosine transform, and an arbitrary block according to the motion vector output from the variable length decoder of the reference frame. An upsampling unit for selecting and inputting a new reference frame and upsampling according to the characteristics of the input frame, and performing motion compensation of the block output from the upsampling unit using the motion vector output from the variable length decoder After that, it is added to the value output from the DC error block output unit and stored as a reference frame. And the output motion compensation and output unit, up-sample the frame used as the reference frame for motion compensation in the spatial domain to obtain 4 × 4 images, thereby reducing hardware complexity and memory bandwidth, and viewing HDTV images on NTSC monitors. There is an effect that can be applied to the card-type products that can be viewed on the set-top box or PC.

Description

Directive's Image Compressor and Its Method

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to digital TV, and more particularly, to an apparatus and method for compressing a video image using motion compensated DCT coding.

Currently, high definition (MP) compression and reconstruction is a standard for US digital TVs, and the MPEG is becoming a standard or a de facto standard in other digital TV markets.

This means that now analog TV, such as NTSC, will gradually be replaced by digital TV.

However, since the price of HDTV, which is the core of digital TV, is quite expensive at the beginning of market formation, SDTV (Standard Definition TV) level demand will exist for a considerable time during the transition from NTSC analog TV to full HDTV.

Here, SDTV does not display the HDTV signal on an HD monitor, but down-converts the HD signal to present a widely available SD monitor, that is, a TV or set-top box shown on an NTSC TV. ). Of course, this SDTV can also receive signals from the SDTV.

For example, it is possible to reduce the 1920 × 1080 60Hz interlaced HD signal to 720 × 480 60Hz interlaced scan, and also to display the 720 × 480 60Hz interlaced SD signal without conversion. Means TV possible.

Currently, the most widely used method of compressing digital TV signals is to use motion compensation in the time domain and two-dimensional transform in the spatial domain.

This is a hybrid method and is called motion coding with motion compensation.

There are several types of two-dimensional methods, among which DCT is the most widely used, and all recent standard video compression methods such as H.261, MPEG-1, MPEG-2, and Grand-alliance methods are all used. It is based on DCT coding that compensates for this movement.

Among the standard video compression methods, the MPEG-2 compliant compression method, that is, a grand-alliance method, is adopted in the United States. The problem with this method is that the signal of the HDTV is not scalable.

Scalable video compression is applicable in large spatial, temporal, signal-to-noise ratio range (SNR domain).

What is needed for SDTV is scalability in the spatial domain, which means that the SD signal is first compressed when the HDTV signal is compressed. The difference signal between the HD signal and the SD signal to make a signal is added to the SD code, which means that the receiver can receive as desired according to its display size.

However, since the signal type of the Grand Alliance is not scalable, the receiver needs to decode the HD signal as determined when displaying the HD signal. However, when displaying the SD level, a special method of downconverting the HD signal to the SD signal is required. Done.

In other words, if the HD signal is displayed on an existing NTSC SD monitor, the size of the image should be reduced.

In this case, the decoder must perform motion compensation and two-dimensional DCT in the reverse order performed by the encoder in order to correctly restore the image.

However, in this case, since the required memory size is required as the original HD video, a method of reducing the memory size in recent years, the DCT coefficients are discarded by Zona filtering, and the IDCT is used as much as desired.

This requires only the size of memory required to be the size of the image to be displayed. In this case, since there are no original pixel values during motion compensation, a motion compensation error accumulates gradually.

A decoder capable of displaying an HD signal on an existing NTSC SD monitor by reducing the size of an image without accumulating the motion compensation error as described above will be described with reference to the accompanying drawings.

1 is a block diagram illustrating a structure of a decoder of a TV receiver performing image compression according to the prior art, wherein a decoder of a TV receiver performing conventional video compression is a DCT coefficient and a motion vector in a compressed and transmitted bitstream. Variable Length Decoding (VLD) unit 11 for decoding and outputting the MVs, and a Zonal Filtering and Inverse Quantization unit 12 for dequantizing the remaining coefficients after the DCT coefficients decoded by the VLD unit 11. The first 4x4 IDCT unit 13 converts and outputs the inverse quantized value by the coarse filtering and inverse quantization unit 12 into a 4x4 block by performing inverse discrete cosine transform, and a frame storing the reference frames. A memory positioning unit 15 and a block positioning unit 15 for outputting four 4x4 blocks according to a motion vector output from the VLD unit 11 among the reference frames output from the frame memory unit 14, 4 output from the block positioning unit 15 The upsampling section 16 for outputting 16 × 16 blocks to four 8 × 8 blocks by performing DCT and IDCT on the 4 × 4 blocks of By using the motion compensation (Motion Compensation: 17) for performing the motion compensation of the 16 × 16 block output from the up-sampling unit 16 using the block, and the motion-compensated block in the MC unit 17 The frame memory unit 14 adds the down sampling unit 18 to reduce and output the 4 × 4 block, and the 4 × 4 blocks output from the down sampling unit 18 and the first 4 × 4 IDCT unit 13, respectively. And an adder 19 for outputting the output and the decoding result.

The upsampling unit 16 receives first 4 × 4 blocks of four 4 × 4 blocks output from the block positioning unit 15 and performs discrete cosine transform to output the first, second, third, and the like. 4 DCTs (161, 162, 163, 164) and the first, second, third, and fourth DCTs (161, 162, 163, 164) are respectively inputted by padding 0 and inverse discrete cosine transform 1, 2, 3, and 4 8 × 8 IDCTs 165, 166, 167, and 168 for outputting an 8 × 8 block.

In addition, the down sampling unit 18 includes an 8 × 8 DCT 181 unit for performing discrete cosine transform on the motion-compensated block in the MC unit 17 to convert it into an 8 × 8 block, and an 8 × 8 DCT ( A coarse filtering unit 182 for outputting a 4x4 block by coarse-filtering the 8 × 8 block outputted from the unit 181; and an inverse discrete cosine transformation for outputting the filtered value of the coarse filtering unit 182. It consists of 2 4x4 IDCT part 183.

Referring to the operation of the decoder of the DVT performing the image compression according to the prior art configured as described above, first, the VLD unit 11 decodes the DCT coefficient and the motion vector in the input compressed bitstream and outputs.

In addition, the pre-join filtering and inverse quantization unit 12 performs the pre-join filtering and inverse quantization of the decoded DCT coefficients output from the VLD unit 11 and outputs the decoded DCT coefficients to the first 4 × 4 IDCT unit 13.

The first 4x4 IDCT unit 13 outputs an error block in which the inverse discrete cosine transform is performed on the inverse quantized value to the adder 21.

In addition, the frame memory unit 15 stores a reference frame having an 8 × 8 block size, and the block positioning unit 15 stores the reference frame stored in the frame memory unit 15 and the motion vector output from the VLD unit 11. And 4 4 × 4 blocks indicated by the motion vector currently input from the reference frame, and are respectively used as the first, second, third, and fourth DCTs 161, 162, 163, and 164 of the upsampling unit 16. Output

The first, second, third, and fourth 4 × 4 DCTs 161, 162, 163, and 164 of the upsampling unit 16 perform discrete cosine transform on the input 4 × 4 blocks, respectively, and then perform the first, second, 3, 4 are output by 8x8 IDCT (165, 166, 167, 168).

The first, second, third, and fourth 8 × 8 IDCTs 165, 166, 167, and 168 pad the 0 into portions other than the input 4 × 4 blocks, respectively, to form 8 × 8 blocks to form the MC unit 17 )

That is, the MC unit 17 inputs 16 × 16 blocks having four 8 × 8 blocks from the upsampling unit 16.

Since the 16 × 16 block input to the MC unit 17 corresponds to the motion vector output from the VLD unit 11, there is no need to scale the motion vector.

Therefore, the MC unit 17 compensates for the motion of the 16 × 16 block output from the upsampling unit 16 using the motion vector output from the VLD unit 11 and outputs the same to the down sampling unit 18.

The 8 × 8 DCT unit 181 of the down sampling unit 18 converts the motion compensated values in the MC unit 17 into 8 × 8 blocks by performing discrete cosine transform, and through the Zonal filter unit 182. Portions other than the 4x4 block are removed and output to the second 4x4 IDCT unit 183.

The second 4x4 IDCT unit 183 performs an inverse discrete cosine transform on the value filtered by the zone filter unit 182 to output the 4x4 block to the adder 19, and the adder 19 is formed by the second unit. The values output from the 1 4x4 IDCT unit 13 and the second 4x4 IDCT unit 183 are added to the frame memory unit 14 and output to the frame memory unit 14 at the same time.

The image compression apparatus of the decoder according to the prior art performs 4 × 4 DCT 4 and 8 × 8 IDCT 4 and 8 × 8 DCT 1 and 4 × 4 IDCT 1 to upsample the complexity of the hardware. Since all 10 high DCTs are performed, this can be difficult to implement directly in hardware.

SUMMARY OF THE INVENTION The present invention has been made to solve the above problems, and an object of the present invention is to provide a video compression apparatus and a method of a dive, which reduces the complexity of hardware.

A feature of the dive image compression device and method thereof according to the present invention is to perform upsampling of a frame used as a reference frame for all motion compensation in a spatial domain.

As described above, the feature of the image compression apparatus of the dive according to the present invention for performing in the spatial domain is to select and input a new reference frame of an arbitrary block according to the motion vector output from the variable length decoder of the reference frame. Up-sampling unit up-sampling according to the characteristics and the motion vector output from the variable-length decoder unit performs motion compensation of the block output from the up-sampling unit and adds it to the DC error block to store as a reference frame And an output motion compensation and output unit.

In addition, another feature of the video compression apparatus of the dive according to the present invention is a block determination unit for outputting an 8 × 8 block according to the motion vector output from the variable length decoder unit of the reference frame, the block determination unit A macroblock classifier for classifying characteristics based on the complexity of the vertical and horizontal directions by receiving the 8 × 8 block output from the interpolator, and an interpolation determiner for determining the order of spatial interpolation according to the values classified by the macroblock classifier; And a spatial interpolation unit outputting the upsampled frame by performing interpolation according to the order determined by the interpolation determining unit.

In addition, the feature of the image compression method of the dive according to the present invention is the step of determining and outputting the 8 × 8 block according to the output motion vector of the up-sampling in the reference frame, and receives the determined 8 × 8 block vertical And classifying the characteristics by the horizontal and horizontal complexity, determining the order of spatial interpolation according to the classified values, and outputting the upsampled frame by performing interpolation according to the determined order. It is possible to lower the memory bandwidth without increasing the memory bandwidth and using the high speed memory when the compensation is processed in the spatial domain.

1 is a block diagram for explaining a configuration of a decoder of a TV unit performing image compression according to the prior art.

FIG. 2 is a block diagram illustrating a configuration of a decoder of a TV receiver performing image compression according to the present invention. FIG.

Explanation of symbols for main parts of the drawings

21: VLD part 22: Jonal filtering and dequantization part

23: 4 × 4 IDCT section 24: frame memory section

25: block determination unit 26: upsampling unit

27: motion compensator 28: adder

261: macroblock classification unit 262: interpolation determination unit

263: Spatial Interpolator

Hereinafter, a description will be given of an image compression apparatus and a method thereof according to the present invention with reference to the accompanying drawings.

FIG. 2 is a block diagram illustrating the structure of a decoder of a DVD which performs image compression according to the present invention. The decoder of the DB according to the present invention is a DCT coefficient and a motion vector (MV) in a compressed and transmitted bitstream. VLD (Variable Length Decoding) unit 21 for decoding and outputting the signals, ZOR filtering and inverse quantization unit 22 for dequantizing the remaining coefficient after the DCT coefficient decoded by the VLD unit 21, and A first 4x4 IDCT unit 23 for performing inverse discrete cosine transform on the inverse quantized value by the filtering and inverse quantization unit 12 to convert into 4x4 blocks, and a frame memory unit for storing reference frames ( 24 and a block determination unit 25 for outputting an 8x8 block according to the motion vector output from the VLD unit 21 of the reference frames output from the frame memory unit 24, and the block determination unit 25. Raise the reference frame of the output 8 × 8 block according to the characteristics Motion Compensation for performing motion compensation of the 16 × 16 block output from the upsampling unit 26 using the upsampling unit 26 for sampling and the motion vector MV output from the VLD unit 21. (Hereinafter referred to as MC) unit 27, the value output from the motion compensator 27 and the value output from the 4x4 IDCT unit 23 are added to the frame memory unit 24 to output and decode the result. It consists of an adder 28 for outputting.

The upsampling unit 26 receives a macroblock classifier 261 that receives 8 × 8 blocks output from the block determiner 25 and classifies characteristics according to vertical and horizontal complexity, and a macroblock classifier ( The interpolation determiner 262 determines the order of spatial interpolation according to the values classified in 261), and the spatial interpolator 263 outputs the upsampled frame by performing interpolation according to the order determined by the interpolation determiner 262. It is composed of

The operation description and the video compression method of the dive image compression device according to the present invention configured as described above will be described.

The operations of the VLD unit 21, the zoned filtering and dequantization unit 22, the 4x4 IDCT unit 23, the frame memory unit 24, and the motion compensation unit 27 of FIG. The same as described with respect to the decoder, and looks at upsampling for performing motion compensation in the spatial domain.

First, the block determiner 25 uses 8x8 macroblocks left and right and up and down from the center of the screen among reference frames corresponding to one screen output from the frame memory unit 24 using the motion vector output from the VLD unit 21. Is output to the macroblock classification unit 261 of the upsampling unit 26.

The macroblock classification unit 261 classifies according to the characteristics of the input macroblock. For example, they are classified by the horizontal and vertical complexity.

Here, the number of classification of macroblocks can be increased or decreased.

That is, the average of the difference between adjacent pixels in each direction can be obtained and classified according to a predetermined reference value, the edge detection result, or the absolute value of the difference between adjacent pixels can also be used.

The determination reference values, which are values used in the classification, may use predetermined values and may be changed according to frames.

The interpolation determiner 262 calculates the order of spatial interpolation according to the value classified by the macroblock classifier 261. Here, the horizontal and vertical directions are obtained separately.

Interpolation is performed according to the spatial interpolation orders obtained in the horizontal and vertical directions to output the upsampled frame to the motion compensator 27.

For example, in the macroblock classifier 261, when the interpolation degree of the values classified in the horizontal direction is 2, the interpolation is performed at 1/2 [1,1], and when 4, 1/8 [1,3,3 , 1] to perform interpolation.

로 연산하고, 4인 경우

로 연산하여 보간을 수행한다.That is, when the interpolation order of the values classified in the horizontal direction in the macroblock classification unit 261 is 2

, If 4,

Interpolation is performed by using.

Here, X1 is a left pixel value of the pixel to be interpolated and X2 is a right pixel value of the pixel to be interpolated. X3 and X4 are first and second pixel values on the left side of the pixel to be interpolated, and X5 and X6 are first and second pixel values on the right side of the pixel to be interpolated.

The order of the spatial interpolation and the coefficient used for interpolation can be changed.

The image compression apparatus and method of the dive according to the present invention can reduce the complexity of the hardware and the memory bandwidth by up-sampling a frame used as a reference frame for motion compensation in a spatial domain to obtain a 4x4 image.

In addition, HDTV video can be applied to a set-top box that can be watched on an NTSC monitor or a card type that can be viewed on a PC.It can be applied to HDTV to show HDTV signals in picture-in-picture or multiple channels simultaneously. There is an effect applicable to multi-channel displays.

Claims (9)

A variable length decoder for decoding and outputting DC coefficients and motion vectors in the compressed and transmitted bitstream; A DC error block output unit for performing Zn filtering on the DC coefficients output from the variable length decoder and performing inverse discrete cosine transform after inverse quantization of the remaining coefficients; An upsampling unit for selecting and inputting a new reference frame of an arbitrary block according to the motion vector output from the variable length decoder unit among the reference frames and upsampling according to the characteristics of the input frame; After performing motion compensation of the block output from the upsampling unit using the motion vector output from the variable length decoder, the motion is stored and output as a reference frame by adding to the value output from the DC error block output unit. The image compression device of the dive characterized in that it is composed of a compensation and an output unit. The apparatus of claim 1, wherein the DC error block output unit is configured to output 4 × 4 blocks by performing inverse discrete cosine transform. The apparatus of claim 1, wherein the upsampling unit is configured to output a new reference frame at an 8 × 8 block in the center of the reference frame. The method of claim 1, wherein the upsampling unit A block determination unit for outputting an 8x8 block according to a motion vector output from the variable length decoder of a reference frame; A macroblock classification unit which receives 8 × 8 blocks output from the block determination unit and classifies characteristics according to vertical and horizontal complexity; An interpolation determination unit for determining the order of spatial interpolation according to the values classified by the macroblock classification unit; And a spatial interpolation unit configured to output an upsampled frame by performing interpolation according to the order determined by the interpolation determining unit. Decoding and outputting the DC coefficient and the motion vector in the compressed and transmitted bitstream, Performing Zn filtering on the DC coefficients output in the above step, inverse quantizing the remaining coefficients, and performing inverse discrete cosine transform to transform and output the arbitrary blocks; Selecting and inputting a new reference frame of an arbitrary block according to the motion vector output in the step of the reference frame and upsampling according to the characteristics of the input frame; And performing a motion compensation of the output block in the upsampling using the motion vector output in the step, and storing and outputting the result as a reference frame by adding to the output value in the converting and outputting step. Directive image compression method. 6. The method of claim 5, wherein the arbitrary blocks are 4x4 blocks in the step of converting and outputting the arbitrary blocks. 6. The method of claim 5, wherein the new reference frame of any block in the upsampling step is an 8x8 block in the center of the reference frame. 6. The method of claim 5, wherein the upsampling is Determining and outputting an arbitrary block according to the output motion vector in a reference frame; Classifying characteristics according to the complexity of the vertical and horizontal directions by receiving the determined 8 × 8 block; Determining an order of spatial interpolation according to the classified value; And outputting the upsampled frame by performing interpolation according to the determined order. The method of claim 8, wherein the determining and outputting of the block comprises: And an 8 × 8 macroblock left and right up and down from the center of the screen among the reference frames corresponding to one screen using the output motion vector.

Images