KR20030006641A - Transcoder and transcoding method therein - Google Patents

Abstract

PURPOSE: A transcoder and a method for transcoding thereof are provided to prevent deterioration of picture quality although video frames are consecutively transcoded. CONSTITUTION: A transcoder includes a VLD(Variable Length Decoder)(111) receiving R1 bit rate video bit stream for entropy decoding, thereby outputting a quantized DCT(Discrete Cosine Transform) coefficient and parameters for motion compensation; a first inverse quantizing part(112) receiving the quantized DCT coefficient for executing an inverse quantization using a quantizing table, and outputting a DCT coefficient including a quantizing error; a first MCD(Motion Compensation Decision)(114) generating a compensation image at a DCT area by executing motion compensation and half pixel motion compensation for the current image at the DCT area by using the motion vector of the VLD and a reference image; an adder(113) generating a restored image at the DCT area by adding the DCT coefficient including the quantizing error and the compensation image; a decoder(110) including a frame memory(115) receiving the restored image to be stored as the reference image to be provided to the first MCD; and an encoder(120) receiving the restored image of the decoder for generating image bit stream having R2 bit rate video bit stream.

Description

Transcoder and transcoding method therein

The present invention relates to a transcoder and a transcoding method thereof, and more particularly, there is no deterioration in image quality even when image frames are continuously transcoded, and a DCT (Discrete Cosine Transform) unit of a conventional transcoder. And a transcoder capable of reducing the complexity of the transcoder by removing an IDCT (Inverse DCT) unit and a transcoding method thereof.

Recently, various methods have been developed to compress and encode video and audio data. In particular, broadcasting stations that produce and broadcast television programs use MPEG technology to compress and encode video data. This MPEG technology has become a de facto standard for transmitting and broadcasting video data.

In connection with such a technique of compressing and encoding data including a series of video and audio equipments using MPEG-2 content to perform digital TV video data streams, the encoding and decoding of various data layers in a digital data transmission stream is performed. need.

Therefore, in order to perform a high compression technique of data including video and audio such as MPEG-2, equipment including an encoder and a transcoder is essential.

Such a basic structure and position of a conventional transcoder is shown in FIG. In the transcoder of FIG. 1, when the original image is input to the encoder 1, the encoder 1 outputs an R1 bitrate video bitstream to the transcoder 4, and the decoder 2 of the transcoder 4 The R1 bitrate video bitstream is received from the encoder 1 and a reconstructed image is output.

The encoder 3 of the transcoder 4 receives the reconstructed image to the decoder 2 and outputs an R2 bitrate video bitstream to the decoder 5, and the decoder 5 outputs the R2 bitrate video. The reconstructed image is output by decoding the bitstream.

Such transcoder previously converts the R1 bitrate video bitstream to another R2 bitrate video bitstream, or converts it to another video format.

FIG. 2 is a block diagram illustrating a conventional transcoder, which includes a variable length decorder (VLD) 11, a first inverse quantization unit 12, and an inverse discrete cosine transform (IDCT) 13.

A decoder unit 10 comprising an adder 14, a motion compensator (MC) 15 and a frame memory 16; A subtractor 21, a DCT 22, a second quantizer 23, a second inverse quantizer 24, an IDCT 25, an adder 26, a frame memory 27 and an MC 28 are provided. It consists of the encoder part 20.

The transcoder operation 1) receives the previous R1 bitrate video bitstream from the VLD 11 of the decoder unit 10, obtains the quantized DCT coefficient value of each video block, and 2) the DCT coefficient value. Obtains the DCT coefficient values of the respective image blocks by the inverse quantization unit 12, 3) obtains the differential image of the reconstructed image in the IDCT 14 using the DCT coefficient values of the respective blocks of the inverse quantization unit 12, 4) The MC 15 acquires a compensation image of the current image using a motion vector of the previous image stored in the frame memory 16 and the current image extracted from the VLD 11, and 5) an adder ( In step 14), the difference image of the IDCT 14 and the compensation image of the MC 15 are summed to obtain a current reconstruction image, and the decoder 10 outputs the current reconstruction image. 6) Then, the current reconstructed video output from the decoder 10 is passed through a DS (Down Sampling) filter 17 to convert the video format.

7) The subtractor 21 of the encoder unit 20 obtains a difference image by subtracting a compensation image obtained by the MC 28 of the encoder unit 20 from the reconstructed image of which the image format of the DS filter 17 is converted. The differential image is converted into discrete cosine transform (DCT) coefficients by the DCT 22 on a block-by-block basis, and 8) the second quantizer 23 bits the differential image converted into discrete cosine transform coefficients by the DCT 22. For rate conversion, quantization is performed using a table different from the quantization table used in the first inverse quantization unit 12 of the decoder unit, and 9) the discrete cosine transform coefficient quantized in the second quantization unit 23 is VLC ( Variable length coding (30) is generated into an R2 bitrate video bitstream different from the R1 bitrate video bitstream input to the original transcoder.

10) Meanwhile, the current image of the second quantization unit 23 is input to the second inverse quantization unit 24 and inverse quantized, and is inverse discrete cosine transformed by IDCT 25 to restore a differential image of the current image. 11) The compensation image of the MC 28 and the difference image are added by the adder 26 to obtain a reconstructed image, and stored in the frame memory 27. The reconstructed image stored in the frame memory 27 is used by the MC 28 to generate a compensation image of the next image.

As described above, the conventional transcoder of FIG. 2 has an advantage that there is no drift phenomenon (a phenomenon in which the image frame of the encoder and the image frame of the decoder are not identical) by using an independent decoder unit and an encoder unit. Due to the complex structure and high computational disadvantages, it was difficult to change the previous R1 bitrate video bitstream to another R2 bitrate video bitstream or to convert to another video format in real time.

FIG. 3 is a block diagram illustrating another conventional transcoder, which is a structure in which a decoder and an encoder are combined, and includes a VLD 31, a first inverse quantizer 32, an adder 33, and a second quantization. A unit 34, a DCT 41, a second inverse quantization unit 36, an adder-subtracter 37, an IDCT 38, an MC 40, a frame memory 40, and a VLC 35 are provided.

The transcoder's operation receives the R1 bitrate video bitstream from the VLD 31, passes through the first inverse quantizer 32, and obtains DCT coefficient values for each block; Receiving a compensation difference image of the MC 40 and the frame memory 40 from the DCT 41 to obtain a DCT coefficient value; The DCT coefficients of the first inverse quantizers 32 and the DCT coefficients of the DCT 41 are summed by the adder 33 to output the DCT coefficients of each block to the second quantizer 34. The two quantization unit 34 obtains the quantized DCT coefficient values of each block and outputs them to the VLC 35; The VLC 35 generates an R2 bitrate video bitstream using the quantized DCT coefficient values.

In addition, a second inverse quantization unit 36 performs quantized DCT coefficient values of each block of the second quantization unit 34 to obtain inverse quantized DCT coefficient values. After obtaining the difference of the DCT coefficient values obtained by the first inverse quantization unit 32, the IDCT 38 obtains the difference image with respect to the current image. Using the motion vector transmitted from the VLD 31, the MC 40 and the frame memory 40 generate a compensation difference image of the current image, and add the difference image to the next image to obtain a difference image for the next image. do.

As described above, the transcoder of FIG. 3 is composed of an encoder and a decoder so that the structure thereof is simple, so that the calculation amount is reduced, and the previous R1 bitrate video bitstream is changed to another R2 bitrate video bitstream, or converted into another video format. While it has the advantage of easy to implement in real time, there is a disadvantage that the degradation of the image quality is intensified because drift occurs.

The present invention has been proposed to solve the problems of the conventional transcoder, and by eliminating the IDCT block in the decoder unit and the DCT block in the encoder, to provide a transcoder capable of real-time transcoding by reducing the amount of calculation The purpose is.

Another object of the present invention is to provide a transcoding method of a transcoder capable of reducing the amount of calculation through the change of the image format in the DCT domain, rather than the image format change at the pixel level.

Another object of the present invention is to provide a transcoding method of a transcoder which improves the quality of a transcoded image by removing a drift phenomenon in an MC block by using the original image as a reference image.

The transcoder for achieving the above object is a VLD 111 that receives an R1 bitrate video bitstream and entropy decodes and outputs parameters for motion compensation such as quantized DCT coefficient values and motion vectors;

A first inverse quantizer 112 which receives the quantized DCT coefficients of the VLD 111, performs inverse quantization using a quantization table, and outputs DCT coefficients including quantization errors;

A first motion compensation decision (MCD) for generating a compensation image in the DCT region by performing motion compensation and half-pixel motion compensation in the DCT region with respect to the current image using the motion vector of the VLD 111 and a reference image. 114,

An adder 113 for adding a compensation image of the first MCD 114 and a DCT coefficient value including a quantization error of the first inverse quantization unit 112 to generate a reconstructed image in the DCT region;

A decoder 110 provided with a frame memory 115 that receives the reconstructed image of the adder 113 and stores it as a reference image for making a compensation image of the next image, and provides the first image to the first MCD 114; ;

The encoder unit 120 is configured to receive a reconstructed image on the DCT region of the decoder unit 110 and generate an R2 bitrate video bitstream.

The transcoding method of the transcoder of the present invention receives an R1 bitrate video bitstream from the VLD 111 of the encoder 120 and entropy decodes the quantized DCT coefficient values and parameters for motion compensation such as a motion vector. Outputting a first step;

A second step of receiving a quantized DCT coefficient value of the VLD 111 from a first inverse quantization unit 112 and performing inverse quantization using a quantization table to output a DCT coefficient value including a quantization error;

The motion vector of the VLD 111 is input from the first MCD 114 to generate a compensation image in the DCT region by performing motion compensation and half-pixel motion compensation in the DCT region with respect to the current image using a reference image. A third step;

A fourth image in which the adder 113 generates and outputs a reconstructed image in the DCT region by adding the compensation image of the first MCD 114 and the DCT coefficient value including the quantization error of the first inverse quantizer 112. Steps;

And a fifth step of receiving a reconstructed image on the DCT region of the adder 113 to generate an R2 bitrate video bitstream.

1 is a block diagram showing the basic configuration of a typical transcoder.

2 is a block diagram showing a conventional transcoder.

3 is a block diagram showing another conventional transcoder.

4 is a block diagram illustrating a transcoder according to the present invention.

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

100: transcoder 110: decoder unit

111: VLD 112: first inverse quantization unit

113, 124: Adder 114, 126: 1st, 2nd MCD

115, 125: Frame memory 116: Image format conversion filter

120: encoder section 121: adder and subtractor

122: second quantization unit 123: second inverse quantization unit

127: VLC

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

4 is a block diagram showing a transcoder of the present invention, wherein the decoder unit 110 has a VLD 111 for receiving an R1 bitrate video bitstream; This VLD 111 is connected to a first inverse quantization unit 112 and a first motion compensation decision (MCD) 114 which outputs a DCT coefficient value including a quantization error; The first inverse quantization unit 112 and the first MCD 114 are connected to the adder 113 and the frame memory is connected to the adder 113 and the first MCD 114. .

In addition, the encoder unit 120 has an adder / subtracter 121 connected to an output terminal of the decoder unit 110 to receive a reconstructed image on the DCT region, which is the output of the decoder unit 110; The adder / subtractor 121 is connected to the output terminal of the second MCD 126; The second MCD 126 is connected to the VLD 111 of the decoder 110 to receive the motion vectors of the decoder 110, and the output terminal of the second MCD 126 is an adder. 124 is connected. Further, another input terminal of the second MCD 126 is connected to the frame memory 125 connected to receive the output of the adder 124.

In addition, an output terminal of the adder / subtracter 121 is connected to a second quantizer 122, and an output terminal of the second quantizer 122 is connected to a second inverse quantizer 123 and a VLC 127. The second inverse quantization unit 123 is configured to be connected to an input terminal of the adder 124.

Here, between the decoder unit 110 and the encoder unit 120, the encoder unit 120 receives a reconstructed image of the adder 113 of the decoder unit 110 and changes the image format in the DCT region. The image format conversion filter 116 may be further provided.

The image format conversion filter 116 may increase or decrease the resolution of the image.

When the decoding operation of the transcoder 100 of the present invention consisting of the decoder 110 and the encoder 120 is described, the R1 bitrate video bitstream is received from the VLD 111 of the encoder 120 and entropy decoded. Parameters for motion compensation such as quantized DCT coefficient values and motion vectors are output.

The first inverse quantization unit 112 receives the quantized DCT coefficients of the VLD 111, performs inverse quantization using a quantization table, and outputs DCT coefficients including quantization errors, and outputs the first MCD 114. In this case, the motion vector of the VLD 111 is input to generate a compensation image in the DCT region by performing motion compensation and half-pixel motion compensation in the DCT region with respect to the current image using a reference image.

The compensation image of the first MCD 114 and the DCT coefficient value including the quantization error of the first inverse quantization unit 112 are added together to generate a reconstructed image in the DCT region by the adder 113 to generate an encoder unit 120. ), The reconstructed image of the adder 113 is supplied from the frame memory 115 and stored as a reference image for making a compensation image of the next image, and then provided to the first MCD 114.

In the encoding operation of the encoder unit 120, the reconstructed image on the DCT region, which is the output of the decoder unit 110, is input from the adder and subtractor 121, and the compensation image for the current image frame is subtracted from the reconstructed image. A differential image is generated in the DCT region.

The VLD 111 of the decoder 110 receives a motion vector from the second MCD 126 and compares reference images to generate a compensation image for the current image frame.

The difference image in the DCT region, which is the output of the adder / subtractor 121, is input to the second quantization unit 122 to output the quantized DCT coefficient value, and the VLC 127 is configured to output the quantized DCT coefficients 122. Transcoding of the transcoder of the present invention is completed by receiving the quantized DCT coefficient value and generating an R2 bitrate video bitstream.

Meanwhile, the quantized DCT coefficient value of the second quantization unit 122 is transferred to the second inverse quantization unit 123 to perform inverse quantization, and the second inverse quantization unit 123 dequantizes the DCT coefficient value. Is transmitted to an input terminal of the adder 124, and the adder 124 sums the compensation image of the current image of the second MCD 126 and the dequantized DCT coefficient value of the second dequantization unit 123, A reference image of an image frame is generated.

The reference image of the next image frame of the adder 124 is transferred to and stored in the frame memory 125, and the reference image is supplied to the second MCD 126 to provide a compensation image for the current image frame. The MCD 126 is to output.

In the MCD, a previous image is used as a reference image, and a compensation image is generated in the DCT region of the current image as a movement vector of the movement of the image in the reference image.

The compensation image in the MCD is generated by the following principle.

The motion compensation in the DCT region in a transcoder operating in the DCT region can be expressed as in Equation 1.

(Equation 1)

Where Prec (t, x, y) is a reconstructed image at x, y coordinates of t frame, e (t, x, y) is a differential image at x, y coordinates, and m _x is the x direction Is a motion vector, and m _y is a motion vector in the y direction,

The motion compensation process in the DCT region uses the windowing and the shifting as much as the moving vector of the image from the previous reference image to display the data corresponding to the size of the macro block (MB) of the image. Movement in the macro block of the video is compensated for.

In addition, MCD compensates for half-pixel motion. Interpolation is necessary for half-pixel motion compensation in the DCT region, and half-pixel motion compensation is performed in the DCT region by obtaining an average between two pixels using linear interpolation.

In order to perform half-pixel motion compensation in the DCT region, the horizontal average in one image block, that is, the interpolation result in the horizontal direction, may be obtained by multiplying the interpolation filter by the original block data.

The transcoder of the present invention configured as described above removes DCT and IDCT of the existing transcoder, compares the current image with the previous reference image in the DCT region, and performs motion compensation and half-pixel motion compensation, thereby By constructing an MCD capable of generating compensation images, the complexity of the transcoder can be eliminated and drift can be prevented.

According to the present invention, a reconstructed image in a DCT region using a compensation image generated by performing motion compensation and half-pixel motion compensation in the first MCD of the decoder unit is generated in the second MCD of the encoder unit, and a reference spirituality is generated. Since the drift is eliminated, there is no deterioration in image quality even if video frames are continuously transcoded, and the complexity of transcoder and the amount of transcoding are reduced by eliminating the DCT and IDCT of the existing transcoder. In addition, the video format can be converted in the DCT area, which can be applied to various applications such as MPEG and H.26x.

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

Claims (8)

A VLD 111 that receives an R1 bitrate video bitstream and entropy decodes and outputs parameters for motion compensation such as quantized DCT coefficient values and motion vectors; A first inverse quantizer 112 which receives the quantized DCT coefficients of the VLD 111, performs inverse quantization using a quantization table, and outputs DCT coefficients including quantization errors; A first MCD 114 for generating a compensation image in the DCT region by performing motion compensation and half-pixel motion compensation in the DCT region with respect to the current image using the motion vector of the VLD 111 and a reference image; An adder 113 for adding a DCT coefficient value including the quantization error of the first inverse quantization unit 112 and a compensation image of the first MCD 114 to generate a reconstructed image in the DCT region; A decoder 110 provided with a frame memory 115 that receives the reconstructed image of the adder 113 and stores it as a reference image for making a compensation image of the next image, and provides the first image to the first MCD 114; ; And an encoder unit (120) for receiving a reconstructed image on the DCT region of the decoder unit (110) and generating an image bitstream having an R2 bitrate video bitstream. The method of claim 1, wherein the encoder unit 120, An adder-subtractor 121 for receiving a reconstructed image on the DCT region of the decoder 110 and subtracting a compensation image for the current image frame from the reconstructed image to generate a difference image in the DCT region; A second MCD 126 that receives the motion vector of the decoder 110 and generates a compensation image for the current image frame; A second quantizer 122 which receives the difference image in the DCT region of the adder and subtractor 121 and outputs a quantized DCT coefficient value; A VLC 127 that receives the quantized DCT coefficient value of the second quantizer 122 and generates an R2 bitrate video bitstream; A second inverse quantizer 123 configured to perform inverse quantization by receiving the quantized DCT coefficient value of the second quantizer 122; An adder 124 for generating a reference image of a next image frame by adding the DCT coefficient value dequantized by the second inverse quantizer 123 and a compensation image of the current image of the second MCD 126; And a frame memory (125) for storing a reference image of the next image frame of the adder (124) and outputting the reference image to the second MCD (126). The reconstruction according to claim 1 or 2, wherein a reconstruction image of the adder 113 of the encoder unit 110 is supplied between the encoder unit 110 and the decoder unit 120 and the image format is changed in the DCT region. And a video format conversion filter (116) for outputting an image to the decoder unit (120). A first step of receiving an R1 bitrate video bitstream from the VLD 111 of the encoder 120 to entropy decode and outputting parameters for motion compensation such as quantized DCT coefficient values and motion vectors; A second step of receiving a quantized DCT coefficient value of the VLD 111 from a first inverse quantization unit 112 and performing inverse quantization using a quantization table to output a DCT coefficient value including a quantization error; The motion vector of the VLD 111 is input from the first MCD 114 to generate a compensation image in the DCT region by performing motion compensation and half-pixel motion compensation in the DCT region with respect to the current image using a reference image. A third step; A fourth step of generating a reconstructed image of the DCT region by the adder 113 by adding the compensation image of the first MCD 114 and the DCT coefficient value including the quantization error of the first inverse quantizer 112. Wow; And a fifth step of receiving a reconstructed image on the DCT region of the adder to generate an R2 bitrate video bitstream. The compensation image of the third step stores the reconstructed image of the DCT region output from the adder 113 as a reference image in the frame memory 115, and the motion vector of the VLD 111 is obtained. When provided to the first MCD 114, a reference image of the frame memory 115 is supplied to the first MCD 114 to provide motion compensation and half pixel motion compensation in contrast to the motion vector and the reference image. Transcoding method of a transcoder, characterized in that the compensation image generated by performing. The method of claim 4, wherein the fifth step, Receiving a reconstructed image on the DCT region from an adder-subtractor 121, and subtracting a compensation image for the current image frame from the reconstructed image to generate a difference image in the DCT region; Generating a compensation image for the current image frame by receiving a motion vector from the VLD (111) and comparing a reference image with a second MCD (126); Outputting a quantized DCT coefficient value by receiving a difference image in a DCT region, which is an output of the adder / subtractor 121, from a second quantizer 122; And receiving the quantized DCT coefficient value of the second quantizer (122) from the VLC (127) to generate an R2 bitrate video bitstream. The method of claim 6, wherein the compensation image of the step of generating the compensation image, The quantized DCT coefficient value of the second quantization unit 122 is transferred to the second inverse quantization unit 123 to perform inverse quantization, and the adder 124 is dequantized by the second inverse quantization unit 123. Transcoding of the transcoder, characterized in that it is generated by adding the DCT coefficient value and the compensation image of the current image of the second MCD 126 and then comparing the reference image of the frame of the image with the motion vector of the VLD 111. Way. The method of claim 4, wherein between the fourth step and the fifth step, Receiving the reconstructed image of the DCT region by the image format conversion filter 116 in the adder 113 and generating a reconstructed image whose image format is changed in the DCT region; The fifth step is a transcoding method of a transcoder, characterized in that for receiving the reconstructed image of the video format is changed in the DCT region generated by the image format conversion filter 116 to generate an R2 bitrate video bitstream.