KR100390813B1 - Apparatus of video decoder for digital television having half-sized hard ware memory and method of the same - Google Patents

Abstract

The present invention provides a video decoding apparatus and method for digital television, which can reduce the memory of the hardware in half and prevent deterioration of image quality through an interpolation method. To this end, the video decoding method of the present invention is a digital television bit. Variable length decoding the stream to separate the motion vector into quantized values and coefficients; Inverse quantization of the coefficient according to the quantization value; A third step of inverse discrete cosine transforming the dequantized coefficients into an image signal of 8 * 8 block units; Adding a video signal of the 8 * 8 block unit and a motion compensated video signal; A fifth step of converting and storing the video signal passed through the fourth step in units of frames; A sixth step of converting the video signal passed through the fifth step into a video signal in a line unit; A seventh step of deducting the magnitude of the video signal passed through the fourth step in half; An eighth step of converting the condensed video signal in units of frames and storing the converted video signal; A ninth step of interpolating the video signal passed through the eighth step to its original size for motion compensation; And a tenth step of repeatedly performing the fourth step by motion compensating the video signal having passed through the ninth step according to the motion vector.

The present invention also provides a digital television video decoding apparatus for such a method.

Description

Apparatus of video decoder for digital television having half-sized hard ware memory and method of the same}

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to video decoders and decoding methods, and more particularly, to digital television video decoding devices and decoding methods in which the hardware memory is reduced by half.

The video compression and multiplexing techniques of the US Digital Television Standard (GA) High Definition Television (HDTV) system are based on the Moving Picture Experts Group (MPEG) -2 standard. There are 24, 30 sequential scan method with 1920 * 1080 pixels and 60㎐ interlaced scan method, not limited to one, and 24㎐, 30㎐ and 60㎐ sequential scan methods with 1280 * 720 pixels Video formats are diverse.

These video formats belong to MPEG-2, which requires a lot of hardware memory to implement a complete digital television decoding device.

In general, digital television receivers do not feel the resolution deterioration greatly with the human eye even when the receiver outputs an image in which the horizontal resolution of the image is reduced by half when the monitor size of the receiver is 50 inches or less.

Therefore, the video decoding apparatus can implement the video decoding apparatus of the simple digital television by reducing the image horizontal resolution to reduce the memory.

1 is a block diagram showing the configuration of a digital television video decoding apparatus according to the prior art.

Referring to FIG. 1, a variable length decoder (VLD) unit 11 for variable length decoding and separating an input digital television bitstream and a variable length decoder unit 11 are provided. An inverse discrete cosine transform having an inverse quantization unit 12 that inversely quantizes an output coefficient and an inverse quantized coefficient that is output from the inverse quantization unit 12 in units of 8 * 8 blocks; An inverse discrete cosine transforming unit 13 for performing IDCT, an adder 14 for adding an image signal output from the inverse discrete cosine converting unit 13, and a motion compensated video signal; The first frame memory unit 15 of 3M bytes for storing and changing the image signal output from the unit 14 in units of frames, and the image signal output from the first frame memory unit 15. Slice buffer (Sli) ce buffer) section 16, the second frame memory section 17 and the variable length decoder section 11 for converting and storing the video signal output from the adder 14 in a frame unit The motion compensation unit 18 is configured to compensate for a video signal output from the second frame memory unit 17 according to a motion vector (MV) and output the motion compensation unit 18 to the adder 14.

The operation of the conventional digital television video decoding apparatus having the above configuration will be described in detail.

When the input digital television bitstream is input to the variable length decoder 11, it is separated into motion vectors having motion information, quantized values, and coefficients through variable length decoding.

The separated motion vector is output to the motion compensator 18, and the quantization value and the coefficient are restored by the inverse discrete cosine transform unit 13 through an inverse discrete cosine transform of 8 * 8 units and then converted into an image signal. It is output to the adder 14.

The adder 14 is predicted by the second frame memory unit 17 through the image signal output from the inverse discrete cosine transforming unit 13 and the motion compensation from the motion compensating unit 18. The signal is added to restore the complete image and output to the first frame memory unit 15.

The first frame memory unit 15 converts an image signal output from the adder 14 into a frame unit and outputs it to the slice buffer unit 16, and the slice buffer unit 16 outputs the first frame. The video signal output from the memory unit 15 is converted in units of lines and finally output.

In addition, the image signal output from the adder 14 is converted and stored in units of frames in the second frame memory unit 17 of 6M bytes, and the motion compensation unit 18 is the variable length decoding apparatus unit. The video signal output from the second frame memory 17 is motion compensated and output to the adder 14 according to the motion vector output from 11.

On the other hand, the conventional digital television video decoding apparatus made as described above has the following problems.

In general, in the case of inverse discrete cosine conversion, which is a method used to reduce the memory in half in a digital television video decoding apparatus, when the block size is 8 * 4, which is half of 8 * 8, the quality of the output image is reduced in proportion to the reduced memory. do.

In addition, since the video decoding apparatus compensates for the motion from the frame memory, the deterioration of image quality becomes more severe because the effects of the image quality deterioration continue to accumulate until the next intra frame comes.

An object of the present invention is to provide a digital television video decoding apparatus and a decoding method capable of preventing the deterioration of an image while reducing the memory in half.

1 is a block diagram showing an example of a digital television video decoding apparatus according to the prior art;

2 is a block diagram showing an example of a digital television video decoding apparatus according to the present invention;

3 is a diagram illustrating a deduction method according to the present invention;

4 illustrates a 5 * 5 Sobel operator according to the present invention;

5A is a diagram illustrating pixels to be interpolated and neighboring pixels;

5B to 5C illustrate interpolation methods of pixels having horizontal and vertical edges according to FIG. 5A;

6A to 6B illustrate a correlation detection algorithm along the horizontal and vertical directions;

7 is a view showing the structure of a median filter according to the present invention;

8 illustrates an algorithm of an interpolation method according to the present invention.

* Description of the main parts of the drawing

20: motion compensation unit

21: variable length decoding unit

22: inverse quantization unit

23: inverse discrete cosine transform unit

24: Adder

25: first frame memory unit

26: slice buffer unit

27: jogger

28: second frame memory unit

29: interpolator

A digital television video decoding method of the present invention for achieving the above object comprises: a first step of variable length decoding a digital television bitstream to separate it into motion vectors, quantization values and coefficients; Inverse quantization of the coefficient according to the quantization value; A third step of inverse discrete cosine transforming the dequantized coefficients into an image signal of 8 * 8 block units; Adding a video signal of the 8 * 8 block unit and a motion compensated video signal; A fifth step of converting and storing the video signal passed through the fourth step in units of frames; A sixth step of converting the video signal passed through the fifth step into a video signal in a line unit; A seventh step of deducting the magnitude of the video signal passed through the fourth step in half; An eighth step of converting the condensed video signal in units of frames and storing the converted video signal; A ninth step of interpolating the video signal passed through the eighth step to its original size for motion compensation; And a tenth step of repeatedly performing the fourth step by motion compensating the video signal having passed through the ninth step according to the motion vector.

In addition, the digital television video decoding apparatus of the present invention comprises: a variable length decoding apparatus unit for inputting a digital television bitstream to output a motion vector, a quantization value, and a coefficient; An inverse quantizer for outputting a dequantized coefficient by inputting the coefficient; An inverse discrete cosine converting unit for outputting an image signal in units of 8 * 8 blocks by inputting the inverse quantized coefficient; An adder configured to add the 8 * 8 block unit video signal and the motion compensated video signal; A first frame memory unit for converting and storing the video signal output from the adder in units of frames; A slice buffer unit which outputs an image signal output from the first frame memory unit in line units; A chopping unit halving the magnitude of the video signal output from the adding unit; A second frame memory unit which converts and stores the video signal output from the carding unit in units of frames; An interpolation unit which interpolates an image signal output from the second frame memory unit to its original size for motion compensation; And a motion compensator for compensating an image signal output from the interpolator according to the motion vector output from the variable length decoding apparatus and outputting the compensation to the adder.

DETAILED DESCRIPTION Hereinafter, exemplary embodiments of the present invention will be described with reference to the accompanying drawings so that those skilled in the art may easily implement the technical idea of the present invention. do.

2 is a block diagram showing a digital television video decoding apparatus according to the present invention.

2, a variable length decoder 21 for outputting a motion vector, a quantization value, and a coefficient as a digital television bitstream as an input, and an inverse quantizer for outputting dequantized coefficients using the coefficient as an input ( 22), an inverse discrete cosine transform unit 23 for outputting an 8 * 8 block unit video signal by inputting the inverse quantized coefficient, and an 8 * 8 block unit video signal and a motion compensated video signal. An adder 24 to be added, a first frame memory 25 to convert and store the video signal output from the adder 24 in units of frames, and to be output from the first frame memory 25. In the slice buffer unit 26 for outputting video signals in units of lines, the extraction unit 27 for halving the magnitude of the image signal output from the adder 24, and the extraction unit 27 Output video signal in frame unit The second frame memory unit 28 converting and storing the interpolation unit 29 and the variable interpolating the video signal output from the second frame memory unit 28 to the original size for motion compensation. And a motion compensator 20 for compensating an image signal output from the interpolator 29 and outputting the compensating image signal to the adder 24 according to the motion vector output from the length decoding apparatus 21.

The digital television video decoding method of the present invention having the above configuration will be described in detail.

First, the variable length decoding of the digital television bitstream is separated into a motion vector, a quantization value, and a coefficient. The coefficient is dequantized according to the quantization value, and the dequantized coefficient is converted into an image signal of 8 * 8 block units. Inverse discrete cosine transform. Subsequently, the 8 * 8 block unit video signal and the motion compensated image signal are added, converted into frame units, stored, converted into line unit video signals, and outputted. Decimation is used to convert and save in frame units.

Subsequently, the deduced video signal is interpolated to the original size for motion compensation, and the motion compensation of the interpolated video signal is repeated according to the motion vector.

Hereinafter, each process will be described in detail with reference to the accompanying drawings.

3 is a diagram illustrating a deduction method according to the present invention.

Referring to FIG. 3, by using an interlaced subsampling method in which pixels to be discarded and pixels to be selected are arranged alternately, odd-numbered pixels are taken in odd-numbered lines and even-numbered lines are discarded by using an interlaced subsampling method. Equation is performed by taking even-numbered pixels and discarding odd-numbered pixels.

The above description will be described in detail.

As the inference filter, a method of substituting a value of a seven-tap Tap induction filter once every two pixels shown in the MPEG rule 11172-2 and a rectangular subsampling method are mainly used.

While the methods are relatively simple and easy to implement, there is a disadvantage in that the pixel information cannot be used vertically, i.e., below or above when interpolating an image.

In particular, in the first method using a filter, the deduced image may be smoother through the filter, but the distortion of the original image is a big problem because the reduced image is not used as it is but a source used for an interpolation filter. To emerge.

In general, the method of deduction can be classified into three types: a method using a FIR filter, a right angle subsampling method, and a mixed subsampling method.

However, the method using the FIR filter has a problem in that pixel values are distorted, and the rectangular subsampling method has a problem in that there are no pixels to be referred to in the vertical direction during interpolation.

Therefore, a half-pixel accurate sub-sampling method can be obtained, and the interpolated sub-sampling method can have pixels to be referred to in the vertical direction during interpolation. By taking the even-numbered pixels and discarding the even-numbered pixels on the even-numbered lines and discarding the odd-numbered pixels, the same amount of data as in the quadrature subsampling method can be reduced and vertical pixels can be used during interpolation.

On the other hand, when using the mixed natural subsampling method, if the input image is an interlaced scan image, a method different from the method in the sequential scan image must be used. This is because the lines to be interpolated and the lines immediately above and below the interlaced images are images of different times. Therefore, when the input image is an interlaced image, different time values are used as reference pixel values when interpolating interlaced subsampled images, and the subsampling method changes. We are tying and sampling the lines of. That is, the odd two pixels may be selected in the first two lines, and the even pixels may be selected in the next two lines.

8 is a block diagram illustrating an algorithm of the interpolation method according to the present invention.

Referring to FIG. 8, the interpolation method according to the present invention detects an edge in a zero padded image input (81), and if there is an edge, interpolates the edge portion according to the direction (82), In the non-edge portion, correlation between neighboring pixels is detected (84). Subsequently, five temporary values corresponding to the correlation are selected at the non-edge portion, and an optimal pixel is selected and interpolated through the median filter (85 and 86), and linear interpolation is performed if there is no correlation. (87).

An interpolation method according to the present invention as described above will be described later in detail with reference to the accompanying drawings.

FIG. 4 is a diagram illustrating an extended 5 * 5 Sobel operator used in edge detection according to the present invention, and FIGS. 5A to 5C illustrate an interpolation method of an edge portion. Drawing.

6A to 6B illustrate a correlation detection algorithm of the present invention, and FIG. 7 illustrates a structure of a median filter.

Referring to FIG. 4, an extended 5 * 5 Sobel operator is used to detect whether an edge of a part to be interpolated in a zero patting image is used. In detail, instead of using an image stored in a memory to find an edge, a zero padded image, that is, an image having a value of "0" in an interpolation portion is used.

In this case, using the 3 * 3 Sobel operator, the pixels used for the calculation to actually find the image are two pixels left and right in the case of the y-operator, and up and down in the case of the x-operator. Only two pixels are used.

Therefore, the 3 * 3 Sobel operator was extended to find an edge using the 5 * 5 Sobel operator.

5A to 5C, FIG. 5A illustrates a pixel X to be interpolated at an edge portion and pixels A, B, C, and D around, and FIG. 5B shows that 'X' is horizontal. 5C shows interpolation in the case of an edge in the direction of) and FIG. 5C shows interpolation in the case where 'X' is an edge in the vertical direction.

Using the 5 * 5 Sobel operator described above, not only can the edge be detected more accurately, but also the pixels stored in the memory, that is, the pixels A, B, C, and D to be interpolated, and the pixel X to be interpolated. It is also possible to detect whether the edge of.

Furthermore, for more accurate interpolation, horizontal and vertical edge maps can be created separately, as shown in FIGS. 5B and 5C, and interpolated according to the direction by looking at whether the 'X' is horizontal or vertical. have.

That is, if four or more of 'A', 'I', 'D', 'D', 'D' are smaller than the threshold, it is considered that there is correlation in the horizontal direction, and the vertical ' If four or more of a ',' b ',' c ',' ㄹ ', and' ㅁ 'are smaller than the threshold, it is considered that there is a vertical correlation.

The algorithms for detecting the correlation in the vertical and horizontal directions are shown in FIGS. 6A to 6B, respectively.

Here, 'A' and 'D' of vertical correlation are values that do not exist, and are used by taking an average of pixel values of both left and right sides, respectively.

In short, the edge region is interpolated by distinguishing the horizontal edge from the vertical edge in the edge region, where the correlation is detected by four horizontal and four vertical pixels in the pixel to be interpolated to select a value to be put into a median filter. The correlation between neighboring pixels is detected using the pixel value.

The median filter is a filter that takes an intermediate value among given N (N is an integer) values, which will be described later with reference to FIG. 7.

Referring to FIG. 7, a median filter is formed by using FIRs on four pixels X1, X2, X3, and X4 in the horizontal and vertical directions, and FIRs on eight pixels. Use

This method uses the FIR filter, which has a filter-specific effect of smoothing the entire image because it has an average effect between the surrounding images. However, if the surrounding pixel values are large, there may be a significant difference from the actual pixel values. Can be.

'Z' represents a pixel to be interpolated, and X ₁ to X ₄ represent horizontal and vertical peripheral pixels X1 to X4.

If there is a horizontal correlation, use the following median filter.

Z = Med {X ₁ , X ₂ , X _HAVE , X _HFIR , X _FIR }

In addition, when there is correlation in the horizontal direction, the following median filter is used.

Z = Med {X ₃ , X ₄ , X _VAVE , X _VFIR , X _FIR }

Here, X _HAVE and X _VAVE are mean values of two horizontal pixels X1 and X2 and average values of two vertical pixels X3 and X4, respectively, and X _HFIR and X _VFIR are four horizontal pixels X1 and X2, respectively. , X13 and X14 are FIR filter values and four vertical pixels X3, X4, X15 and X16 are FIR filter values. In addition, X _FIR is a value in which _FIR is applied to the surrounding twelve pixels X1 to X12.

That is, if there is a horizontal directionality in relation to neighboring pixels, the horizontal left and right pixels (X1, X2) values (X ₁ , X ₂ ), the average values (X _HAVE ) of the X ₁ and X ₂ , and the horizontal directions 4 a pixel value (X _FIR) covered with the FIR values covered with a FIR filter (X _HFIR) and around twelve pixels (X1 ~ X12) to (X1, X2, X15, X16 ) are enclosed in the median filter.

In addition, if there is a vertical orientation in the association with the surrounding pixels, vertical vertical pixel (X3, X4) values (X ₃ , X ₄ ), average values (X _VAVE ) of the X ₃ and X ₄ , and horizontal direction four pixel (X3, X4, X13, X14) to the FIR filter the capped value (X _VFIR) and around twelve pixels (X1 ~ X12) bars which thrust the value (X _FIR) covered the FIR to the median filter, pixels In order to reduce the error of the values, the pixel values of the median filter are adaptively changed according to the correlation between the neighboring pixels.

Accordingly, the interpolation with the neighboring pixels may be recognized in advance, and the interpolation pixels having a high correlation may be included in the median filter, so that more accurate interpolation may be performed.

The digital television video decoding apparatus and decoding method of the present invention made as described above is performed by inferring and storing the inverse discrete cosine transformed video signal to half of its original size before storing it in the frame memory. In order to minimize the deterioration of the image quality of the output image by performing interpolation when the memory is read, it has been found through the embodiment that a satisfactory output image can be obtained while reducing the memory to half of the prior art.

Although the technical idea of the present invention has been described in detail according to the above preferred embodiment, it should be noted that the above-described embodiment is for the purpose of description and not of limitation. In addition, those skilled in the art will understand that various embodiments are possible within the scope of the technical idea of the present invention.

According to the present invention as described above, it is possible to obtain a satisfactory output image while reducing the memory to half of the prior art, thereby reducing the cost.

Claims (8)

A digital television video decoding apparatus, A variable length decoder for outputting a motion vector, a quantization value, and a coefficient as a digital television bitstream as an input; An inverse quantizer for outputting a dequantized coefficient by inputting the coefficient; An inverse discrete cosine converting unit for outputting an image signal in units of 8 * 8 blocks by inputting the inverse quantized coefficient; An adder configured to add the 8 * 8 block unit video signal and the motion compensated video signal; A first frame memory unit for converting and storing the video signal output from the adder in units of frames; A slice buffer unit which outputs an image signal output from the first frame memory unit in line units; A chopping unit halving the magnitude of the video signal output from the adding unit; A second frame memory unit which converts and stores the video signal output from the carding unit in units of frames; An interpolation unit which interpolates an image signal output from the second frame memory unit to its original size for motion compensation; And A motion compensator for compensating an image signal output from the interpolator according to the motion vector output from the variable length decoder and outputting the compensation to the adder. Digital television video decoding apparatus comprising a. A digital television video decoding method, Variable length decoding the digital television bitstream into motion vectors and quantized values and coefficients; Inverse quantization of the coefficient according to the quantization value; A third step of inverse discrete cosine transforming the dequantized coefficients into an image signal of 8 * 8 block units; Adding a video signal of the 8 * 8 block unit and a motion compensated video signal; A fifth step of converting and storing the video signal passed through the fourth step in units of frames; A sixth step of converting the video signal passed through the fifth step into a video signal in a line unit; A seventh step of deducting the magnitude of the video signal passed through the fourth step in half; An eighth step of converting the condensed video signal in units of frames and storing the converted video signal; A ninth step of interpolating the video signal passed through the eighth step to its original size for motion compensation; And A tenth step of repeatedly performing the fourth step by motion compensating the video signal having passed through the ninth step according to the motion vector; Digital television video decoding method comprising a. The method of claim 2, Inference of the seventh step, Digital television video decoding method characterized by using a interwoven subsampling method. The method of claim 3, wherein The interwoven subsampling method, An odd line takes pixels of an odd number of rows and discards even rows, and an even line takes an even number of pixels and discards odd rows of pixels. The method of claim 2, The interpolation of the ninth step, Detecting an edge and interpolating the edge portion according to the direction; Detecting correlations between neighboring pixels at portions other than the edges; And Selecting five temporary values corresponding to the correlation at the non-edge portion, and selecting an optimal pixel through a median filter to interpolate Digital television video decoding method comprising a. The method of claim 5, The edge detection is, A digital television video decoding method comprising detecting whether an edge of a part to be interpolated in a zero-patting image is edged by using an extended 5 * 5 Sobel operator. The method according to claim 5 or 6, Interpolating the edge portion, And using the result of claim 6, to interpolate the edge region by distinguishing a horizontal edge from a vertical edge in an edge region. The method of claim 5, Detecting the correlation, In order to select a value to be inserted into the median filter, a correlation between neighboring pixels is detected using four horizontal and four vertical pixel values in the pixel to be interpolated. Digital TV video decoding method.