KR20070105534A - Method and apparatus for reduction mpeg noise using wavelet transform - Google Patents

Abstract

A method and an apparatus for reducing MPEG(Moving Picture Experts Group) noise by using wavelet transform are provided to set up a specific block boundary as a filtering area and apply one-dimensional filter to the filtering area, thereby effectively removing blocking noise, shortening a filtering time, and reducing image blurring. A method for reducing MPEG(Moving Picture Experts Group) noise by using wavelet transform comprises the following steps of: obtaining an edge map separating a block boundary from an input image(S410); setting up a predetermined number of pixels as a basic filtering area around the block boundary and resetting up an adaptive filtering area according to the obtained edge map(S420,S440); and performing one-dimensional LPF(Low Pass Filtering) to the adaptive filtering area in a horizontal direction and a vertical direction to remove the MPEG noise(S470).

Description

Method and apparatus for MPEG noise removal using wavelet transform {Method and apparatus for reduction MPEG noise using wavelet transform}

1 is a block diagram illustrating an encoder and a decoder for processing a video that is compressed and transmitted by a conventional MPEG technology.

2 is a screen showing an example of blocking noise.

3 is a block diagram of a DTV block for explaining an MPEG noise removing apparatus using the wavelet transform according to the present invention.

Figure 4 is a block diagram showing an MPEG noise removal apparatus using a wavelet transform in accordance with the present invention.

5 is a control flowchart illustrating an MPEG noise removing method using a wavelet transform according to the present invention.

6 is a diagram illustrating a state of setting a specific filtering region of a block boundary portion on a screen.

FIG. 7 is a diagram illustrating a state of performing filtering on a specific filtering region of FIG. 6. FIG.

8 is a block diagram illustrating a filter configuration for implementing the mallet wavelet transform of the present invention.

9A to 9B are diagrams for explaining the threshold holding process.

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

100 tuner section

120: Transport Demux

130: decoder

134: display processing module

160: control unit

210: edge map detection means

220: adaptive filtering region setting means

230: low pass filtering means

The present invention relates to an MPEG noise removing method, and more particularly, to an MPEG noise removing method and apparatus using a wavelet transform for effectively removing MPEG noise, particularly blocking noise, generated when receiving a DTV.

In general, a video display device is a device for displaying image information by using a screen, and representative devices include a television system for receiving a radio wave of a broadcasting station and displaying a video, and a hard disk built in a set-top box or a television body. And a personal video recorder (PVR) for recording and playing back information.

In conventional analog television broadcasting, video signals to be transmitted are modulated by AM or FM and transmitted through radio waves or cables. With the recent development of digital technologies such as digital image compression and digital modulation and demodulation, the standardization of digital television (DTV) broadcasting is rapidly progressing, and existing terrestrial, satellite and cable broadcasting are based on MPEG (Moving Picture Experts Group). Is digitizing.

As such, digital broadcasting can provide higher quality of service than analog service according to the development of digital video / audio compression technology and digital transmission technology, and can transmit multiple broadcast programs in the same bandwidth, digital communication media and digital storage. There is an advantage in that interoperability with the media can be improved.

In general, digital broadcasting compresses and transmits video and audio data using MPEG technology.

1 is a block diagram illustrating an encoder and a decoder for processing a video that is compressed and transmitted using a conventional MPEG technology.

The encoder 10 includes a SIF preprocessor 11, a motion estimation unit 12, a DCT unit 13, a quantization unit 14, and a variable length encoder 15.

The SIF preprocessing unit 11 converts the SIF (Source Input Format) signal by selecting a specific field for the original image data and limiting the band of the luminance and chrominance signals as the processing prior to compression.

The motion estimation unit 12 performs a motion estimation by calculating a motion vector with respect to the input image in units of macroblocks. Prediction methods based on motion vectors include an I picture (Intra Picture), a P picture (Predictive Picture), and a B picture (Bidirectionally Predictive Picture).

The DCT unit 13 then performs a two-dimensional discrete cosine transform (DCT) on the difference between the motion compensation image and the input image into a block size of 8x8 pixels.

The quantization unit 14 quantizes the DCT transform coefficients using a matrix table in consideration of visual characteristics.

The variable length coding unit 15 compresses the quantized value read while zigzag scanning with variable length coding (VLC) by a combination of a run length code and a Huffman code.

The decoder 20 is composed of a variable length code decoder 21, an inverse quantization unit 22, an IDCT unit 23, a motion compensator 24, and a post processor 25.

The variable length code decoder 21 decodes the input variable length code (VLD: Variable Length Decoding) to obtain a quantization coefficient and a motion vector.

The inverse quantization unit 22 returns the quantization coefficients to the DCT coefficients by the inverse quantizer (multiplication by the Q value).

The IDCT unit 23 calculates respective pixel values (luminance and color difference) for each 8x8 pixel block by an inverse DCT (Inverse DCT). The pixel value here becomes the actual pixel value in the I picture and the difference value between the corresponding pixel value in the P picture and the B picture.

The motion compensator 24 decodes the P picture or the B picture by adding the pixel values of the block compensated by the difference value and the motion vector (MC).

The post processor 25 de-interlaces the decoded screen to implement the original image. That is, since the MPEG encoded video is one field, the number of horizontal scanning lines is half that of a general TV. Therefore, when the decoded screen is finally displayed on the TV, it is possible to display the same field screen repeatedly or to display them in a progressive scan method, but usually the line interpolation processing is performed on the decoder side.

The digital broadcast is compressed and received by MPEG-2, and the noise generated thereby is mainly MPEG noise (quantization noise) generated by quantization. In addition, the MPEG noise includes a blocking artifact noise and a ringing artifact noise. In particular, the most noticeable noise among the MPEG noises is blocking noise that appears in a block boundary portion of a flat area. .

However, the conventional method for removing noise removes MPEG noise generated when receiving a DTV broadcast by using a filter designed for overall noise reduction, that is, dispersion. That is, conventionally, "the variance of the noise is much smaller than the variability of the pixels included in the detail or edge region, and the distribution of the variability of the flat region is similar to the distribution of the noise variability" Under two assumptions, the filtering coefficients are determined by determining the variance of the signal and noise, and the filtering is performed by applying the filtering coefficients over the entire area.

Therefore, the conventional MPEG noise removal method is not effective to remove the blocking noise appearing in the MPEG boundary, especially the block boundary part, by performing the normal filtering method over the entire area as described above, and performs the entire area filtering. As a result, excessive processing time is required, and an image blurring phenomenon occurs.

Accordingly, the present invention is to solve the above-mentioned conventional problems, by setting a specific block boundary portion as the filtering region and applying a one-dimensional filter to effectively remove the blocking noise and to shorten the filtering execution time. It is an object of the present invention to provide a method and apparatus for removing MPEG noise using a wavelet transform that reduces image blurring.

According to an aspect of the present invention, there is provided a method for removing noise from an MPEG, comprising: a) obtaining an edge map for dividing a block boundary surface from an input image; b) setting a predetermined number of pixels as a basic filtering area around the block boundary plane, and resetting an adaptive filtering area according to the edge map obtained in the step; c) removing MPEG noise by performing one-dimensional (1D) low pass filtering (LPF) in the horizontal and vertical directions on the adaptive filtering region.

The edge map is preferably obtained using a two-dimensional Mallet wavelet transform function that extracts localized high frequency components. The basic filtering area sets six pixels as the basic filtering area around the block boundary plane. The adaptive filtering region may determine the closest edge pixels left and right around the block boundary by using the left and right edge maps of blocks horizontally adjacent to the block boundary surface, and then reference the edge pixels closer to the block boundary surface among the determined edge pixels. Decide on The low pass filtering uses a Gaussian filter composed of LPF {1,1,2,4,2,1,1}, but the filter coefficient weight of the sub-pixel determined as an edge in the edge map is zero. Set it.

In addition, the MPEG noise removing apparatus according to the present invention comprises: edge map detection means for obtaining an edge map for dividing a block boundary surface from an input image; Adaptive filtering region setting means for setting a predetermined number of pixels around the block boundary plane as a basic filtering region and resetting the adaptive filtering region according to the edge map obtained in the step; And low pass filtering means for removing MPEG noise by performing one-dimensional (1D) low pass filtering (LPF) in the horizontal and vertical directions on the adaptive filtering region.

The edge map detection means; It is desirable to obtain the edge map using a two-dimensional Mallet wavelet transform function that extracts localized high frequency components. The adaptive filtering region setting means; Six pixels around the block boundary plane are set as the basic filtering area. The adaptive filtering region setting means; After determining the closest edge pixels left and right around the block boundary using the left and right edge maps of blocks horizontally adjacent to the block boundary surface, the adaptive filtering area is determined based on the edge pixels closer to the block boundary surface among the determined edge pixels. Decide The low pass filtering means uses a Gaussian filter composed of LPF {1,1,2,4,2,1,1}, and uses a filter coefficient weight of less than or equal to a pixel determined as an edge in the edge map. Set to 0 to perform low pass filtering.

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

3 is a block diagram of a DTV block for explaining an MPEG noise removing apparatus using a wavelet transform according to the present invention.

As shown, the DTV for explaining the present invention, a DTV tuner unit 100 for physically selecting a specific channel and receiving a digital broadcast signal in the form of a transport stream in the selected channel, the DTV tuner unit 100 A demodulator unit 110 for demodulating the received transport stream; A transport demux 120 for demultiplexing the demodulated transport stream into respective video bitstreams (MPEG-2 Bitstream), audio bitstreams (AC-3 Bitstream), and PSIP data; A decoder unit 130 for decoding the demultiplexed signals from the transport demux 120 and outputting video and audio signals; A display unit 140 displaying the video signal on a screen; A speaker 150 for outputting the audio signal; Outputs a channel and program selection signal to the DTV tuner unit 100 and the demodulator unit 110 according to the user's program selection and the PSIP data output from the transport demux 120, and displays an on-screen display (OSD). A control unit 160 that provides a function, determines a reservation recording time set by a user, and outputs a storage command of an image signal; The storage unit 170 stores a video signal received through the DTV tuner 100 according to the reservation recording time determination of the controller 160 and a storage command of the video signal.

In addition, the decoder 130 may include a video decoding module 131 for decoding a video bitstream (MPEG-2 Bitstream) into a video signal, and an audio decoding module for decoding an audio bitstream (AC-3 Bitstream) into an audio signal. 132 and receiving time stamps from the respective video and audio bitstreams to process clock recovery and A / V synchronization signals of the video decoding module 131 and the audio decoding module 132. A clock recovery and A / V synchronization processor 133; And a display processing module 134 for converting the video signal decoded by the video decoding module 131 into a displayable image output format.

Figure 4 is a block diagram showing an MPEG noise removal apparatus using a wavelet transform according to the present invention.

The MPEG noise removing apparatus according to the present invention, as shown, the edge map detection means for obtaining an edge map (edge map) for dividing the block boundary surface from the input image; Adaptive filtering region setting means (220) for setting a predetermined number of pixels around the block boundary plane as a basic filtering region and resetting the adaptive filtering region according to the edge map obtained in the step; And low pass filtering means 230 for performing one-dimensional (1D) low pass filtering (LPF) on the adaptive filtering region in horizontal and vertical directions to remove MPEG noise.

The edge map detection means 210; It is preferable to obtain the edge map by using a two-dimensional Mallet wavelet transform function that extracts localized high frequency components.

The adaptive filtering region setting means (220); Six pixels around the block boundary plane are set as the basic filtering area. The adaptive filtering region setting means (220); After determining the closest edge pixels to the left and right around the block boundary using the left and right edge maps of blocks horizontally adjacent to the block boundary surface, the adaptive filtering region is determined based on the edge pixels closer to the block boundary surface among the determined edge pixels. Decide

The low pass filtering means 230 uses a Gaussian filter composed of LPF {1,1,2,4,2,1,1}, and includes a sub-pixel filter determined as an edge in the edge map. Low pass filtering is performed by setting the coefficient weight to zero.

5 is a control flowchart illustrating an MPEG noise removing method using a wavelet transform according to the present invention.

As shown, the present invention may include obtaining an edge map for dividing a block boundary plane from an input image (S410); Setting a predetermined number of pixels as a basic filtering area around the block boundary plane, and resetting an adaptive filtering area according to the edge map obtained in the step (S420 to S440); And removing MPEG noise by performing one-dimensional (1D) low pass filtering (LPF) in the horizontal and vertical directions on the adaptive filtering region (S450 to S470).

The edge map is preferably obtained using a two-dimensional Mallet wavelet transform function that extracts localized high frequency components (S410). In the basic filtering region, 6 pixels are set as the basic filtering region around the block boundary plane (S420). The adaptive filtering area may determine an edge pixel closest to the left and right around the block boundary using the left and right edge maps of blocks horizontally adjacent to the block boundary surface (S430), and then the edge closer to the block boundary surface among the determined edge pixels. The determination is made based on the pixel (S440). The low pass filtering uses a Gaussian filter composed of LPF {1,1,2,4,2,1,1} (S450), and filter coefficient weights of sub-pixels determined to be edges in the edge map. Set to 0 (S460).

Referring to the preferred embodiment of the present invention configured as described above are as follows.

First, the present invention performs an operation of obtaining an edge map for dividing a block boundary plane from an input image (S410).

Accordingly, the edge map detection unit 210 obtains an edge map that separates the block boundary plane from the input image. At this time, the edge map detection means 210; The edge map is obtained using a two-dimensional Mallet wavelet transform function that extracts localized high frequency components.

The Wavelet Transform theory is a subband coding technique used in the field of multi-resolution signal processing and compression of voice or image, which has been studied in computer vision since it was first introduced in applied mathematics. It is a kind of signal interpretation method related to the back light.

Such methods of observing signals at various resolutions and dividing them into various frequency bands have been tried in engineering and physics. Wavelet transform is a signal analysis method that decomposes a signal using a set of basis functions, such as a Fourier transform. However, in the wavelet transform, the wavelet is a base function having a finite length unlike the base functions of the Fourier transform. In the wavelet transform, these basis functions are called wavelets and correspond to a kind of bandpass filter.

Wavelet transformation is achieved by a set of wavelets obtained by the contraction and expansion of a circular wavelet called a mother wavelet. Methods of extracting local frequency components from a signal include a short time fourier transform (STFT) and a wavelet transform (WT) that interpret a frequency component of a signal in a window by covering a signal with a window.

In particular, the present invention utilizes a two-dimensional (2D) mallet wavelet transform function that exhibits good performance for detecting abrupt changes in the signal. The two-dimensional mallet wavelet transform function performs a wavelet transform on the input image, detects the position and size of the local peak (edge) of the wavelet transformed image, and encodes the low bit rate and high It is a way to realize efficient encoding.

Thereafter, a predetermined number of pixels are set as a basic filtering area around the block boundary plane, and an adaptive filtering area is reset according to the edge map obtained in the step (S420 to S440).

Accordingly, the adaptive filtering region setting unit 220 sets a predetermined number of pixels around the block boundary plane as the basic filtering region, and resets the adaptive filtering region according to the edge map obtained in the step.

At this time, the adaptive filtering region setting means 220; Six pixels around the block boundary plane are set as the basic filtering area. The adaptive filtering region setting means (220); After determining the closest edge pixels to the left and right around the block boundary using the left and right edge maps of blocks horizontally adjacent to the block boundary surface (S430), the adaptability is based on the edge pixels closer to the block boundary surface among the determined edge pixels. The filtering area is determined (S440).

FIG. 6 is a diagram illustrating a state of setting a specific filtering region of a block boundary portion on a screen, and FIG. 7 is a diagram illustrating a state of performing filtering on a specific filtering region of FIG. 6.

As shown in the figure, the left and right edge pixels are determined based on the block boundary by using the left and right edge maps of blocks horizontally adjacent to the block boundary, and then among the determined edge pixels, the edge pixels closer to the block boundary surface are used as a reference. Determine the filtering area. Here, 1D LPF is performed in the horizontal and vertical directions on the block boundary filtering region set adaptively to the edge. In this case, a Gaussian filter composed of LPF {1,1,2,4,2,1,1} is used, but it is adaptive to the edge by setting the filter coefficient weight below the pixel determined to be an edge in the edge map to 0. As the coefficients change, filtering removes MPEG noise. That is, in the illustrated figure, the coefficient values of the filter determined by the left and right three edge maps are {0,0,2,4,2,1,0}.

Subsequently, one-dimensional (1D) low pass filtering (LPF) is performed on the adaptive filtering region in horizontal and vertical directions to remove MPEG noise (S450 to S470).

At this time, the low pass filtering means 230 removes MPEG noise by performing one-dimensional (1D) low pass filtering (LPF) in the horizontal and vertical directions on the adaptive filtering region.

The low pass filtering means 230 uses a Gaussian filter composed of LPF {1,1,2,4,2,1,1}, and includes a sub-pixel filter determined as an edge in the edge map. Low pass filtering is performed by setting the coefficient weight to zero.

8 is a block diagram illustrating a filter configuration for implementing a two-dimensional mallet wavelet transform of the present invention, and FIGS. 9A to 9B are diagrams for describing a threshold holding process.

As shown, we apply a lowpass filter and a highpass filter on each row of the two-dimensional image to find the wavelet coefficients on the first scale, and four subband images LL, LH, HL that are the same size as the four original images. , HH is produced.

Since the image we use requires edge information to obtain the edge map, we use the information of the HH subband including both the horizontal and vertical edge information.

Here, LL is subsampled to 2 by applying low pass filters 310 and 320 in the horizontal and vertical directions to the original image, and HL is applied to the high pass filter 330 in the vertical direction. Contains error components. LH is a high-pass filter 340 is applied in the horizontal direction and contains an error component of the horizontal frequency. And, HH is a high pass filter 350, 360 is applied in the horizontal and vertical direction. Thereafter, the threshold holding unit 370 threshes the wavelet coefficients obtained above to obtain an edge map.

9A to 9B are diagrams for explaining a threshold holding process, and a gray level threshing is performed on an original image as shown in FIG. 9A to obtain an image image as shown in FIG. 9D.

As described above, the present invention sets a specific block boundary portion as a filtering area and applies a one-dimensional filter to effectively remove blocking noise, shorten the filtering time, and reduce image blurring.

The present invention described above can be applied not only to flat panel display devices such as PDPs and LCDs, but also to projection TVs, general TVs, monitors, and other video devices, and can be equally applied to analog TVs, digital TVs, and satellite TVs. In addition, it can of course be applied to all NTSC, PAL, SECAM method.

Although the preferred embodiments of the present invention have been illustrated and described above, the present invention is not limited to the above-described embodiments, and the present invention is not limited to the above-described embodiments without departing from the spirit of the present invention as claimed in the claims. Anyone skilled in the art can, of course, make various changes and make such changes as fall within the stated claims.

As described above, according to the MPEG noise removing method and apparatus according to the present invention, the block noise is effectively removed by setting a specific block boundary portion as a filtering region and applying a one-dimensional filter, thereby shortening the filtering execution time. In addition, there is an effect of reducing the image blurring phenomenon.

Claims (10)

Obtaining an edge map for dividing the block boundary plane from the input image; Resetting an adaptive filtering area according to the edge map obtained in the step of setting a predetermined number of pixels around the block boundary plane as a basic filtering area; And removing MPEG noise by performing one-dimensional (1D) low pass filtering (LPF) in the horizontal and vertical directions on the adaptive filtering region. The method of claim 1, The edge map is obtained by using a two-dimensional Mallet wavelet transform (Mallat wavelet transform) function for extracting local high-frequency components, MPEG noise cancellation method using a wavelet transform. The method of claim 1, The basic filtering region, the MPEG noise removing method using a wavelet transform, characterized in that the six pixels are set as the basic filtering region around the block boundary plane. The method of claim 1, The adaptive filtering region may determine the closest edge pixels left and right around the block boundary by using the left and right edge maps of blocks horizontally adjacent to the block boundary surface, and then reference the edge pixels closer to the block boundary surface among the determined edge pixels. MPEG noise removal method using a wavelet transform, characterized in that determined by. The method of claim 1, The low pass filtering is A Gaussian filter composed of LPF {1,1,2,4,2,1,1} is used, and a filter coefficient weight of less than a pixel determined as an edge in the edge map is set to 0. MPEG noise cancellation using wavelet transform. Edge map detection means for obtaining an edge map for dividing a block boundary plane from an input image; Adaptive filtering region setting means for setting a predetermined number of pixels around the block boundary plane as a basic filtering region and resetting the adaptive filtering region according to the edge map obtained in the step; And low-pass filtering means for removing MPEG noise by performing one-dimensional (1D) low-pass filtering (LPF) in the horizontal and vertical directions on the adaptive filtering region. The method of claim 6, The edge map detection means; Apparatus for removing MPEG noise using a wavelet transform, wherein the edge map is obtained using a two-dimensional Mallet wavelet transform function for extracting localized high frequency components. The method of claim 6, The adaptive filtering region setting means; The MPEG noise removing apparatus using the wavelet transform, characterized in that 6 pixels are set as the basic filtering area around the block boundary surface. The method of claim 6, The adaptive filtering region setting means; After determining the closest edge pixels to the left and right around the block boundary using the left and right edge maps of blocks horizontally adjacent to the block boundary surface, the adaptive filtering region is determined based on the edge pixels closer to the block boundary surface among the determined edge pixels. MPEG noise removal apparatus using a wavelet transform characterized in that it determines. The method of claim 6, The low pass filtering means, Low pass filtering by using a Gaussian filter composed of LPF {1,1,2,4,2,1,1}, and setting the filter coefficient weight below the pixel determined as an edge in the edge map to 0 MPEG noise removal apparatus using a wavelet transform, characterized in that for performing.

Images