Methods and apparatus for removing blocking effect in a motion picture decoder

The present invention relates to a method and apparatus for removing blocking effect caused by quantization error in a motion picture decoder.

In general, the amount of data associated with visual information is so large that its storage would require enormous storage capacity. Although the capacities of utilizing several storage media are substantial, the access speeds are usually inversely proportional to the capacity. Storage and transmission of such data require large capacity and bandwidth. To eliminate the need for large storage capacity, there is an image data compression technique, which reduces the number of bits required to store or transmit image without any appreciable loss of data.

The image data compression removes redundancies contained in image signals. The redundancies comprises a spectral redundancy among colors, a temporal redundancy between successive screens, a spatial redundancy between adjacent pixels within the screen, and a statistical redundancy. Here, a method of image coding for removing the spatial redundancy is transform coding, which divides original input images into small size blocks and processes them individually.

In the transmitter, each blocks of original image is converted by the transform coding and transform coefficients are generated. The transform coefficients are quantized and transmitted to the receiver. In the receiver, the transforn, coefficients are inverse quantized and converted so that each blocks of original image is regenerated.

Fig. 1 shows a block diagram of a conventional digital motion picture coder/decoder, which is widely used in image processing system such as a High Definition Television (HDTV).

In Fig. 1, the motion picture coder comprises a differential image generator (DIG), a discrete cosine transform unit (DCT unit), a quantizer (Q), a variable length coding unit (VLC unit), an inverse quantizer (IQ), an inverse discrete cosine transform unit (IDCT unit), an adder (ADD), a frame memory, a motion estimator, and a motion compensator.

In DIG, a current image and a predicted image are inputted, and a differential image is generated. The generated differential image is outputted to the DCT unit to be divided into blocks. The DCT unit processes every block into DCT coefficients. The DCT coefficients are then quantized according to a quantization step size in the quantizer. The quantized coefficients are coded according to the Huffman Table in the VLC unit. The quantized coded coefficients are then transmitted to the channel.

The predicted image inputted to the DIG is obtained as following. First, quantized DCT coefficients from the quantizer are quantized inversely in the inverse quantizer. The inverse quantized DCT coefficients are converted to image data in the IDCT. The converted image data are inputted to the adder. In the adder, original images are regenerated by using the transformed image data and previous image data from the motion compensator. The regenerated images from the adder are stored in the frame memory. From the frame memory, the previous images are outputted by delaying them in frame units. In the motion estimator, the previous image signals from the frame memory and the current image signals are compared for difference between the two frames, and a motion vector is generated. In the motion compensator, the predicted image having a pixel value similar to the current frame is outputted by shifting the previous image signal outputted from the frame memory as much as the motion vector.

In the motion picture coder like the above, intra-mode frames are coded and transmitted. However, in the case of inter-mode frames, differential signals obtained through the motion estimation and the motion compensation should be coded and transmitted in order to decrease the transmission rate. To solve the above problem, a switch is disposed between the DIG and the motion compensator. Accordingly, the switch is opened when the intra-frames are inputted, and the switch is closed when the inter-frames are inputted.

The motion picture decoder comprises a variable length decoding unit (VLD unit), an inverse quantizer, an inverse discrete cosine transform unit (IDCT unit), an adder, a frame memory, and a motion compensator.

In the motion picture decoder, the input image signals are decoded by the VLD unit. The decoded signals are then quantized inversely by the inverse quantizer. The inverse quantized image data are converted inversely by the IDCT unit to be outputted to the adder. The image data from the adder are stored and delayed in the frame memory. The consequent stored delayed data from the frame memory are outputted to the motion compensator in order to be compensated with reference to the previous images. The compensated signals are then outputted to the adder.

The intra-mode frames and inter-mode frames are regenerated according to the switch disposed between the adder and motion compensator. Namely, in case of the intra-mode frames, the output data from the IDCT unit are directly outputted to the adder. However, in case of the inter-mode frames, the output data from the IDCT are added to the previous image data from the motion compensator and the resulting data are transmitted to the adder to regenerate a current image signal.

In the decoded digital images like the above, blocking effect occurs near to discontinuous boundary between blocks. The occurrence of this blocking effect is generated during the transform coding process of the divided blocks of digital images. Further, when the quantization step size is expanded during quantization, the quantization error increases and the blocking effect in the discontinuous boundary between blocks becomes even more apparent.

US-A-5,367,385 discloses a method and apparatus for processing block coded image data to reduce boundary artifacts between adjacent image blocks.

In view of the foregoing, it is an object of the present invention to provide methods and apparatuses for removing blocking effect due to a quantization error in a motion picture decoder having loss to the original image data.

In order to achieve the above object, the present invention provides a method for removing blocking effect in a motion picture decoder comprising the steps of: <ul id="ul0001" list-style="none" compact="compact"><li>a) extracting boundary pixel values of a current block and each adjacent block to said current block, which exist between boundaries of said current block and said each adjacent block;</li><li>b) extracting difference values between boundary pixel values of said current block and boundary pixel values of said adjacent block, and calculating the mean value for the difference values; and</li><li>c) limiting the mean value between -1/2 quantization step size and 1/2 quantization step size;</li></ul>    characterised by comprising a step d) of adding the limited mean value to said each pixel value of said current block, and outputting corrected images, the step d) being carried out after the step c).

According to a second aspect of the present invention there is provided an apparatus for removing blocking effect in a motion picture decoder comprising: <ul id="ul0002" list-style="none"><li>a) a frame memory for receiving and storing decoded image signals in frame units and a corrected feedback current block;</li><li>b) a current block boundary pixel extracting means for extracting sequentially boundary pixel values of said current block stored in the frame memory;</li><li>c) an adjacent block boundary extracting means for extracting sequentially boundary pixel values of adjacent blocks to said current block stored in the frame memory;</li><li>d) a mean value calculating means for calculating a mean value for difference values between boundary pixel values of said current block and of said adjacent blocks, and outputting a mean value; and</li><li>e) a mean value limiting means for limiting the mean value between -1/2 quantization step size and 1/2 quantization step size, and outputting the limited mean value;    characterised by comprising:</li><li>f) a current block extracting means for extracting a current block stored in the frame memory so as to output pixel values of said current block; and</li><li>g) an adding means for adding the mean value from said mean value limiting means to each pixel value of said current block from said current block extracting means so as to output the corrected current block before feeding back said corrected current block to said frame memory.</li></ul>

The above and other objects, features, and advantages of the present invention will be apparent from the following detailed description of the preferred embodiments of the invention in conjunction with the accompanying drawings, in which: <ul id="ul0003" list-style="none"><li>Fig. 1 is a block diagram illustrating a conventional motion picture coder and decoder;</li><li>Figs. 2A to 2D are views illustrating loss of an information due to a quantization error;</li><li>Fig. 3 is a block diagram illustrating a motion picture decoder comprising an apparatus for removing blocking effect in accordance with the present invention;</li><li>Fig. 4 is a block diagram illustrating an apparatus for removing blocking effect in accordance with a first preferred embodiment of the present invention;</li><li>Fig. 5 is a view illustrating transformed 8 x 8 pixels of current block and upper, lower, left, right adjacent blocks;</li></ul>

Reference will now be made in detail to the present invention, examples of which are illustrated in the accompanying drawings. Wherever possible, the same reference numbers will be used throughout the drawings to refer to the same or like parts.

Figs. 2A to 2D show a process of an information loss due to a quantization error. When a DCT coefficient is quantized according to a quantization step size, the quantized DCT coefficient is expressed by QF= [ <maths id="math0001" num=""><math display="inline"><mrow><mfrac><mrow><mtext mathvariant="italic">F</mtext></mrow><mrow><mtext mathvariant="italic">step</mtext></mrow></mfrac></mrow></math><img file="EP0808068B1_D0001.tif" /></maths> + <maths id="math0002" num=""><math display="inline"><mrow><mfrac><mrow><mtext>1</mtext></mrow><mrow><mtext>2</mtext></mrow></mfrac></mrow></math><img file="EP0808068B1_D0002.tif" /></maths> ] ( [ ] : Gaussian symbol ). For example, in Fig. 2B, the quantized DCT coefficient value is QF=1, if the range of the DCT coefficient F is <maths id="math0003" num=""><math display="inline"><mrow><mfrac><mrow><mtext>1</mtext></mrow><mrow><mtext>2</mtext></mrow></mfrac></mrow></math><img file="EP0808068B1_D0003.tif" /></maths> X step ≤ F &lt; <maths id="math0004" num=""><math display="inline"><mrow><mfrac><mrow><mtext>3</mtext></mrow><mrow><mtext>2</mtext></mrow></mfrac></mrow></math><img file="EP0808068B1_D0004.tif" /></maths> X step. As shown in Fig.2D, the inversely quantized coefficient value F' is determined by expression F'=1 X step when the above quantized coefficient value QF=1 is quantized inversely. Accordingly, the difference (| F-F' |) between the original DCT coefficient value F and the inversely quantized DCT coefficient F' may be <maths id="math0005" num=""><math display="inline"><mrow><mfrac><mrow><mtext>1</mtext></mrow><mrow><mtext>2</mtext></mrow></mfrac></mrow></math><img file="EP0808068B1_D0005.tif" /></maths> X step at its maximum. Namely, the quantization error that which cause the blocking effect in the digital image falls within 0 ≤| F-F' | ≤ <maths id="math0006" num=""><math display="inline"><mrow><mfrac><mrow><mtext>1</mtext></mrow><mrow><mtext>2</mtext></mrow></mfrac></mrow></math><img file="EP0808068B1_D0006.tif" /></maths> X step.

Fig. 3 shows a motion picture decoder comprising an apparatus for removing blocking effect <b>36</b> of the present invention, which is connected to an end portion of the decoder in Fig. 1. Decoded image signals and a quantized step size information are inputted into the apparatus for removing blocking effect <b>36</b> such that blocking effects in the image signals are removed and the corrected image signals are outputted.

Fig. 4 shows a first embodiment of the apparatus for removing blocking effect <b>36</b>. The apparatus for removing blocking effect <b>36</b> comprises a frame memory <b>40</b>; a current block boundary pixel extractor <b>41</b>; a adjacent block boundary pixel extractor <b>42</b>; a mean value calculator <b>43</b>; a mean value limiter <b>44</b>; a current block extractor <b>45</b>; and an adder <b>46</b>. The mean value limiter <b>44</b> comprises a multiply <b>44-1</b> for multiplying quantized step size by <maths id="math0007" num=""><math display="inline"><mrow><mfrac><mrow><mtext>1</mtext></mrow><mrow><mtext>2</mtext></mrow></mfrac></mrow></math><img file="EP0808068B1_D0007.tif" /></maths>; a selector <b>44-2</b> for selecting and outputting a minimum <b>MIN</b>; a multiply <b>44-3</b> for multiplying the quantized step size by - <maths id="math0008" num=""><math display="inline"><mrow><mfrac><mrow><mtext>1</mtext></mrow><mrow><mtext>2</mtext></mrow></mfrac></mrow></math><img file="EP0808068B1_D0008.tif" /></maths> ; and a selector <b>44-4</b> for selecting and outputting a maximum <b>MAX</b>.

Fig. 5 shows 8 × 8 pixel current block, adjacent upper, lower, left, and right blocks and boundary pixels between each block. Here, the current block for removal of blocking effect is illustrated. The boundary pixels <b>a</b>_<b>0</b><b>-a</b>_<b>7</b> of the current block correspond to the boundary pixels <b>A</b>_<b>0</b><b>-A</b>_<b>7</b> of the upper adjacent block <b>A.</b> The boundary pixels <b>b</b>_<b>0</b><b>-b</b>_<b>7</b> of the current block correspond to the boundary pixels <b>B</b>_<b>0</b><b>-B</b>_<b>7</b> of the left adjacent block <b>B.</b> The boundary pixels <b>c</b>_<b>0</b><b>-c</b>_<b>7</b> of the current block correspond to the boundary pixels <b>C</b>_<b>0</b><b>-C</b>_<b>7</b> of the lower adjacent block <b>C</b>. The boundary pixels <b>d</b>_<b>0</b><b>-d</b>_<b>7</b> of the current block correspond to the boundary pixels <b>D</b>_<b>0</b><b>-D</b>_<b>7</b> of the right adjacent block <b>D.</b>

In Fig. 4, the decoded image signals are inputted and stored in the frame memory <b>40</b> in frame units. In the current block boundary pixel extractor <b>41</b>, boundary pixels values ℓ of the current block, which lies between the current block and the adjacent blocks are extracted from the frame memory <b>40.</b> Namely, 32 boundary pixel values <b>a</b>_<b>0</b><b>-a</b>_<b>7</b><b>, b</b>_<b>0</b><b>-b</b>_<b>7</b><b>, c</b>_<b>0</b><b>-c</b>_<b>7</b><b>, d</b>_<b>0</b><b>-d</b>_<b>7</b> are outputted sequentially from the current block boundary pixel extractor <b>41.</b>

In addition, by the adjacent block boundary pixel extractor <b>42</b>, boundary pixels values m of the adjacent blocks, which lies between the current block and the adjacent blocks are extracted from the frame memory <b>40.</b> Namely, 32 boundary pixel values <b>A</b>_<b>0</b><b>-A</b>_<b>7</b><b>, B</b>_<b>0</b><b>-B</b>_<b>7</b><b>, C</b>_<b>0</b><b>-C</b>_<b>7</b><b>, D</b>_<b>0</b><b>-D</b>_<b>7</b> are outputted sequentially from the adjacent block boundary pixel extractor <b>42.</b>

In the mean value calculator <b>43,</b> the pixel values ℓ of the current block and the pixel values <b>m</b> of the adjacent blocks are sequentially inputted. Each difference value ℓ-<b>m</b> between the inputted pixel values is calculated by the mean value calculator <b>43</b>. Further, a mean value for each difference value is calculated and ouputted from the mean value calculator <b>43</b>. The mean value is obtained by the following expression.<maths id="math0009" num=""><img file="EP0808068B1_D0009.tif" /></maths>

The above equation 1 is applied when there are 4 adjacent blocks. However. when there are 2 or 3 adjacent blocks, the mean value is obtained by dividing all the added boundary pixel values by 24 or 16 respectively.

In the mean value limiter <b>44</b>, the mean value and a quantization step size are inputted. The inputted mean value is limited between the minimum - <maths id="math0010" num=""><math display="inline"><mrow><mfrac><mrow><mtext>1</mtext></mrow><mrow><mtext>2</mtext></mrow></mfrac></mrow></math><img file="EP0808068B1_D0010.tif" /></maths> × step and maximum <maths id="math0011" num=""><math display="inline"><mrow><mfrac><mrow><mtext>1</mtext></mrow><mrow><mtext>2</mtext></mrow></mfrac></mrow></math><img file="EP0808068B1_D0011.tif" /></maths> × step by the mean value limiter <b>44</b>, and the limited mean value <b>LIM</b> is outputted. Namely, in the minimum selector <b>44-2</b>, the mean value is compared to maximum <maths id="math0012" num=""><math display="inline"><mrow><mfrac><mrow><mtext>1</mtext></mrow><mrow><mtext>2</mtext></mrow></mfrac></mrow></math><img file="EP0808068B1_D0012.tif" /></maths> × step to select a smaller value. In the maximum selector <b>44-4,</b> the smaller value obtained from the minimum selector <b>44-2</b> is compared to the minimum - <maths id="math0013" num=""><math display="inline"><mrow><mfrac><mrow><mtext>1</mtext></mrow><mrow><mtext>2</mtext></mrow></mfrac></mrow></math><img file="EP0808068B1_D0013.tif" /></maths> × step to select a bigger value. The limited mean value <b>LIM</b> from the maximum selector <b>44-4</b> is outputted to the adder <b>46.</b>

In the current block extractor <b>45,</b> the current block is extracted from the frame memory <b>40</b>. 64 pixel values of the extracted current block are sequentially outputted to the adder <b>46</b>.

In the adder <b>46</b>, the limited mean value <b>LIM</b> from the mean value limiter <b>44</b> is added to each pixel value of the current block outputted from the current block extractor <b>45</b>. The added and corrected current block is inputted into a display processor (not shown). In addition, the corrected current block is fed back to the frame memory <b>40</b>. In the frame memory <b>40</b>, the stored current block is replaced with the corrected currect block from the adder <b>46</b>. This corrected current block is used for removal of blocking effect in the next block.

Another method for calculating the mean value utilizes a threshold level <b>TH</b>. TH prevents a damage in edges of original image during the removal process of blocking effect. To obtain the mean value <b>M</b>, the threshold level is determined and an absolute value from the difference value between boundary pixels smaller than the threshold value is selected. For example, when a 256 level image is coded/decoded, the threshold value is obtained as the following. A range of quantization error falls within -4 to 4 when the quantization step size 8 is utilized, and a range of the absolute difference between boundaries of blocks can be predicted to be 0 to 8. Further, a mean absolute difference between the block boundary is determined to be 4 if the quantization error is an uniform distribution. Accordingly, the threshold level greater than 4 is selected since the threshold value must be greater than the differences of the boundary pixel values. Namely, when the determined threshold level is 4 and the absolute value |ℓi - m_i | of differences between the boundary pixel values of less than or equal to 4 is selected to yield a mean value <b>M</b>, At this time, the mean value <b>M</b> is obtained by the following expression.<maths id="math0014" num=""><math display="block"><mrow><mtext>Equation 2.</mtext><mspace linebreak="newline" /><msub><mrow><mtext> if (|ℓ</mtext></mrow><mrow><mtext>i</mtext></mrow></msub><msub><mrow><mtext> - m</mtext></mrow><mrow><mtext>i</mtext></mrow></msub><msub><mrow><mtext> | ≤ TH) {S = S + ( ℓ</mtext></mrow><mrow><mtext>i</mtext></mrow></msub><msub><mrow><mtext> - m</mtext></mrow><mrow><mtext>i</mtext></mrow></msub><mtext> ) C = C + 1}</mtext><mspace linebreak="newline" /><mtext> M = </mtext><mfrac><mrow><mtext mathvariant="italic">S</mtext></mrow><mrow><mtext mathvariant="italic">C</mtext></mrow></mfrac></mrow></math><img file="EP0808068B1_D0014.tif" /></maths>    (S and C are initialized to 0 at each block)

The obtained mean value is inputted into the mean value limiter <b>44</b> for limiting the mean value within the range of quantization error.

While this invention has been described in connection with what is presently considered to be the most practical and preferred embodiment, it is to be understood that the invention is not limited to the disclosed embodiment, but, on the contrary, it is intended to cover various modifications and equivalent arrangements included within the scope of the appended claims.