Signal adaptive filtering metod, signal adaptive filter and computer readable medium for storing program therefor

ES 2 328 439 B2

DESCRIPTION

Image data filtering procedure.

Technical field

The present invention is a divisional application of the Spanish patent application No. 20050077, for "Procedure and device for adaptive filtering of image data". The invention relates to data filtering, and more particularly to an image data filtering method to reduce the block effect.

Previous technique

In general, the standards for encoding images such as MPEG of the International Standardization Organization (ISO) and the H.263 protocol recommended by the International Telecommunication Unit (ITU), adopt the estimate block-based motion and discrete cosine transform (DCT) blocks. When an image is greatly compressed, block-based encoding can lead to the well-known block effect. A typical block effect is grating noise in a homogeneous area in which adjacent pixels have relatively similar pixel values. Another block effect is the stair-step noise that is generated along the edges of the image. Also, a hum noise is a typical Gibb phenomenon that is produced by truncation when the DCT coefficients are quantized to greatly compress the image.

The grating noise shows traces of the block-based procedure at the edges between blocks when the compressed data is displayed on a screen after restoration. Consequently, the edges between the blocks can be identified. Also, ladder noise has a stepped shape at the edges of the image, so that an uneven edge can be seen in the image. Also, overlapping of images with a predetermined interval can be seen due to hum noise.

Various methods have been proposed to reduce the block effect and hum noise that occurs when block-based coding is performed. In accordance with the encoding according to the H.261 protocol, a simple 3x3 Low Pass Filter (LPF) is used as a loop filter to reduce the block effect [“Video Codee for Audiovisual Serviees at Px62 kbit / s”, CCITT Recommendation H.261, December 14, 1990]. Also, a simple edge loop filter has been proposed to reduce block effect and hum noise [G. Bjontegaard, “A Simple Edge Loop Filter to Reduce Blocking and Mosquito Noise”, ISO / IEO JTC1 / Sc29 / WG11 MPEG96 / 0617, January 1996, and “A Simple Edge Loop Filter to Reduce Blocking and Mosquito Noise”, ITU SC15 LBC Expert Group ITU-LBC-96-032, January 1996]. The edge loop filter outputs linear values ​​from two pixels adjacent to the block boundary and replaces the two pixel values ​​with the linear values. Such an edge loop filter can reduce the block effect, but not the hum noise. In order to reduce such hum noise, a non-linear filter using a binary index has been proposed [Y.Itoh, “Detail Preserving Nonlinear Filter using Binary Index,” ISO / IEC JTC1 / SC29 / WG11 MPEG95 / 0357, November 1995]. However, the non-linear filter cannot reduce the block effect.

Presentation of the invention

In order to solve the problems mentioned above, an objective of the present invention consists, in particular, in providing a method of filtering image data to reduce the block effect, which comprises obtaining the absolute value of the difference between the values of pixels; comparing the absolute value of the difference obtained with a quantization parameter (Q); and determining whether the pixel values ​​are modified according to the result of the comparison when filtering is performed on the pixels.

In general, the method comprises inverse quantization of the image data; generating a flag information indicating whether the image data requires filtering, based on the inversely quantized image data and; filtering the image data according to the generated flag information.

The invention can provide a method of signal adaptive filtering, capable of reducing the block effect and hum noise of image data when a frame is composed of blocks of a predetermined size, the method comprising the following steps: (a) generate information about the block to reduce the effect of the same and information about the hum to reduce the hum noise, from predetermined pixel coefficients of the upper and left limit areas of the data block when a frame obtained by decomposing image data into a bit string for inverse quantization is an intraframe; and (b) adaptively filtering the image data that goes through the inverse quantization and inverse discrete cosine transform based on the information generated about the block and hum noise.

Preferably, step (a) further comprises the step of setting the information about the block and the buzz noise of the previous frame that corresponds to a motion vector as information about the block and the buzz noise of the current frame if said frame is an interframe, and set the hum noise information to "1" which represents the image data that needs to be filtered if there is a residual signal of the current block inversely quantized, and the information about the block and the hum noise is determined according to the coefficients of a

ES 2 328 439 B2 pixel A arranged in the upper left corner of the block, a pixel B arranged to the right of pixel A and a pixel C arranged below pixel A.

Preferably, the information about the block is composed of both horizontal and vertical information, the first being set to "1" which represents the image data that needs to be filtered only when the coefficient of pixel A is not equal to "0", or any coefficient of the pixels in the left boundary zone of the block is not equal to “0”, and the vertical information on the block being set to "1" which means the image data that needs to be filtered only when the coefficient of pixel A is not equal to "0" or any coefficient of the pixels of the upper limit zone is not equals "0", and the hum information is set to "1" which means the image data needs to be filtered when any coefficient of pixels other than pixels A, B or C in the block is not equal to "0".

Preferably, the information about the block is composed of both horizontal and vertical information, the first being set to "1" which means the image data that needs to be filtered when all the coefficients of pixels A, B and C of the block are not. equal to "0" or any coefficient of the pixels in the left boundary zone is not equal to "0", and the vertical information on the block being set to "1" which means the image data that needs to be filtered when all the coefficients of pixels A, B and C are not equal to "0" or any coefficient of the pixels of the The upper limit zone of the block is not equal to "0", and the information about the hum is fixed at "1" which means the image data that needs to be filtered when any coefficient of the pixels other than pixels A, B and C of the block is not equal to "0".

Preferably, in order to reduce the block effect, in step (b), the horizontal (or vertical) filtering is performed using a weighted filter that presents a predetermined weighted value when the information about the horizontal (or vertical) block ) of the block is equal to “1” and the information about the hum is equal to “0”, and when the information about the horizontal (or vertical) block of the block is not equal to “1” or the information about the hum is not is equal to "0", A comparison is made between an absolute value of the difference between adjacent pixels and a Q value used as a dividend to quantize the block, and then filtering is carried out with a predetermined value according to the result of the comparison.

The invention can also provide a signal adaptive filter capable of reducing the block effect and hum noise of image data when a frame is composed of blocks of a predetermined size, comprising: a mode flag checking unit for checking a flag in order to determine whether a frame is an intraframe or an interframe when decomposing image data from a bitstream for inverse quantization; an information generator about the "intra" filtering to generate information about the block to reduce the block effect and information about the hum to reduce it, from predetermined pixel coefficients of the upper and left limit areas of the block of data when it is determined by the flag verification unit that the frame is an intra-frame; an information generator about the "inter" filtering to set the information about the block and the information about the buzz of the previous frame that corresponds to a motion vector as information about the block and the buzz of the current frame if said frame is an interframe , and setting the hum information to "1" if there is a residual signal from the current inversely quantized block; and an adaptive filter for adaptively filtering image data passing through an inverse quantizer and an inverse direct cosine transformer based on the block and hum information generated by the "intra" filtering information generators and the filtering "inter".

The invention can be carried out with a general purpose digital computer that executes a program from a computer-readable medium, including but not limited to storage media such as magnetic storage media (e.g. ROM, floppy disks, disks). hard, etc.), optically readable media (for example, CD-ROMs, DVDs, etc.) and carrier waves (for example, Internet transmissions). Accordingly, the present invention can be carried out as a means capable of being computer processed.

The invention can also provide a computer-readable medium containing a computer program for adaptive signal filtering capable of reducing the block effect and buzz noise of image data when a frame is composed of blocks of a predetermined size, wherein the adaptive signal filtering comprises the following steps: (a) generate information about the block to reduce the effect of the block and information about the hum to reduce the noise of the hum, from predetermined pixel coefficients of the upper and left boundary areas of the data block when a frame obtained by decomposing image data into a bit string for inverse quantization is an intraframe; (b) set the information about the block and the buzz of the previous frame that corresponds to a motion vector as information about the block and the buzz noise of the current frame if said frame is an interframe, and set the information about the hum in "1" representing the image data that needs to be filtered if there is a residual signal of the current block inversely quantized; (c) adaptively filtering the image data that passes through the block-based inverse quantization and inverse discrete cosine transform, based on the information generated about the block and hum.

Also, a computer-readable medium containing a computer program can be provided for a method for filtering hum noise that occurs when block-based compressed image data is decoded, wherein said method comprises the following steps: (a) performing a gradient operation on the block undergoing inverse quantization and inverse discrete cosine transform using predetermined one-dimensional horizontal and vertical gradient operators; (b) generate a binary edge map that

ES 2 328 439 B2 represents whether each pixel is an edge pixel, using an absolute value of the difference between the operated gradient value of a pixel and the value of the adjacent pixel, and a Q value used as a dividend to quantize the block; and (c) performing filtering by applying a predetermined filtering window to the generated edge binary map.

Brief description of the drawings

Figure 1 is a block diagram of a signal adaptive filter for reducing block effect and hum noise in accordance with the present invention;

Figure 2 is a flow chart illustrating a procedure for signal adaptive filtering;

Figure 3 shows an inverse quantized block having 8x8 pixels;

Figure 4 is a flow chart illustrating the step of generating information that is used to filter an intraframe;

Figure 5 is a flow chart illustrating the step of generating information that is used to filter an interframe;

Figure 6 shows the arrangement of pixels adjacent to the boundary of the block, in order to illustrate filtering to reduce the block effect; Y

Figure 7 shows the arrangement of pixels to be processed in the current block.

Best mode to carry out the invention

In Figure 1, a signal adaptive filter for reducing block effect and hum noise in accordance with the present invention includes a mode flag verification unit 120, an "inflate" filtering information generator 130, a "inter" filtering information generator 140 and an adaptive filter unit 150. When the image data is decomposed into a bit string for inverse quantization, the mode flag checking unit 120 checks whether the frame is an intraframe or an interframe. The "inflate" filtering information generator 130 generates block information to reduce block effect and hum information from predetermined pixels of the upper and left boundary areas of the data block when determined by means of the flag verification unit 120 so that the frame is an intraframe. When said mode flag verification unit 120 determines that the frame is an interframe, the "inter" filtering information generator 140 generates information about the block and the buzz of the previous frame that corresponds to a motion vector as block information. and hum of the current frame. In the event that there is a residual signal of the inversely quantized current block, the hum is set to "1". Adaptive filtering unit 150 adaptively filters block image data that has passed through an inverse quantizer (Q ¹ ) 100 and a discrete cosine inverse transformer (DCT ¹) 110, according to the information about the block and the hum generated by the information generator about it. “inflate” filtering 130 and the “inter” filtering information generator 140.

Next, a procedure for signal adaptive filtering will be described. Figure 2 is a flow chart illustrating the procedure for signal adaptive filtering in accordance with the present invention. The bitstream image data encoded by an encoder is decoded by a decoder to be reproduced. To this end, the data is decomposed into a bit string and then inverse quantized using the inverse quantizer 100 (step 200). In this case, the image data is made up of a plurality of frames, each frame being made up of a plurality of blocks. Figure 3 shows an inverse quantized block having 8x8 pixels that make up the frame.

Before filtering the frame data using an inverse discrete cosine transform (IDCT), a flag is checked to determine whether the frame is an intraframe or an interframe (step 210). In the case that the frame is an intraframe (step 220), the information used to filter the intraframe is generated (step 230). In the case that the frame is an interframe, the information used to filter the interframe is generated (step 240). Next, the frame data that has passed through the IDCT 110 is adaptively filtered according to the general filtering information, thereby eliminating the block effect and hum noise (step 250).

Figure 4 is a flow chart illustrating in detail the step of generating information used to filter the intra-frame. As can be seen in Figure 4, in the case that it is determined by means of the flag verification unit so that the frame is an intra-frame, the coefficient of pixel A of Figure 3 is verified (step 400) . In the case that only the coefficient of pixel A is not equal to "0", the horizontal (HBI) and vertical (VBI) block information is set to "1" (step 410). In the event that any of the coefficients of the pixels (8 pixels, including pixels A and B) that belong to the upper limit zone 300 of the block indicated in Figure 3 is not equal to "0" (step 420 ), the VBI is set to "1" (step 430). Otherwise, the VBI is set to "0" (step 440). Likewise, if any coefficient of the pixels (8 pixels, including pixels A and C) that belong to the zone of the left limit 310 of the block shown in Figure 3 is not equal to "0" (step 450), set the HBI at "1" (step 460). Otherwise, the HBI is set to "0" + (step 470).

ES 2 328 439 B2

Once the HBI and VBI are set, the hum information (RI) used to filter the hum noise is generated. That is, if any coefficient of pixels other than pixels A, B and C of the block shown in Figure 3 is not equal to "0" (step 480), the IR is set to "1" (step 490). Otherwise, the IR is set to "0" (step 495). In this case, the HBI and VBI are set to "1" only when the coefficient of pixel A is not equal to "0" (step 400). However, when the HBI and VBI are set to "1", even if all the coefficients of pixels A, B and C are not equal to "0", a favorable effect can be obtained to some extent when performed at then an adaptive filtering to signals.

Figure 5 is a flow chart illustrating the step of generating information used to filter the interframe. In the case where it is determined by the flag verification unit 120 that the frame is an interframe, the HBI, VBI and RI of the intraframe are transferred to the HBI, VBI and RI of the interframe according to the motion vector (step 500). Also, in the case that there is a residual signal after motion compensation (step 510), the RI is updated (step 520).

When the information about the block and the hum for filtering is generated in the manner described above, filtering is performed adaptively according to said information. First, a filtering procedure to reduce the block effect will be described. Filtering to reduce the block effect is classified as horizontal and vertical filtering. In this case, an explanation of horizontal filtering will be provided. Figure 6 shows the arrangement of pixels adjacent to the boundary of the block, in order to illustrate filtering to reduce the block effect. It is determined if the HBI and RI of blocks I and J of Figure 6 are equal to "0" and, if so, a weighted filtering is carried out on pixels A, B, C, D, E and F of Figure 6 using a 7-stop low-pass filter (LPF) (1,1,1,2,1,1,1).

In the case that the HBI and the RI of the blocks I and J of Figure 6 are not equal to "0", a filtering is carried out on the pixels B, C, D and E using the following algorithm.

d = D-C;

If (ABS (d) <Q) {

D = D - (d / 2); d = ED;

If (ABS (d) <Q) d = CB;

if (ABS (d) <Q)

C = C + (d / 2);

Ε = E - (d / 4);

Β = B + (d / 4);

} or {if (ABS (d / 2) <2Q)

If (d> 0) {

D = D - (Q-ABS (d / 2));

C = C + (Q-ABS (d / 2));

} or {+ (Q-ABS (d / 2));

+ (Q - ABS (d / 2));

} d = ED;

ES 2 328 439 B2 if (ABS (d) <Q) d = CB;

if (ABS (d) <Q)}

E = E- (d / 4);

Β = B - (d / 4);

In the algorithm included above, ABS represents an absolute value and Q represents a dividend used when the blocks that make up the frame are quantized.

In detail, if the absolute value (ABS (d)) of the difference (d) between pixels D and C is equal to or less than Q, the current pixel value of pixel D is set by subtracting d / 2 from the current value pixel, and the current pixel value of pixel C is set by adding d / 2 to the current pixel value. Likewise, if the absolute value (ABS (d)) of the difference (d) between pixels E and D is equal to or less than Q, the current pixel value of the pixels is set by subtracting d / 4 from the current pixel value . Also, if the value. Absolute (ABS (d)) of the difference (d) between pixels C and B is equal to or less than Q, the current pixel values ​​of pixel B are set by subtracting d / 4 from the current pixel value. In the same way as before, if the pixel values ​​of the pixels B, C, D and E are set according to the algorithm different from the previous one, which is evident to those skilled in the art, for which reason the explanation of the same. Also, vertical filtering is carried out according to the same principle as horizontal filtering.

Next, a procedure for reducing hum noise will be described. First, the information generated about the drone is verified. In the case that the hum information is set to "1", filtering is carried out. Otherwise, no filtering is done. For this purpose, the pixels at the edge of the block that have undergone inverse quantization and IDCT are determined. In order to determine the edge pixels, a gradient operation is performed on the blocks that have undergone inverse quantization and IDCT using one-dimensional horizontal and vertical gradient operators.

Next, an absolute value of the difference between a gradient operation pixel value and the value of the adjacent pixel, and a Q value used as a dividend when the block is quantized, are used to generate a binary edge map representing the edge of each pixel. In this case, the block is 8X8 pixels, and the size of the edge binary map is represented as a two-dimensional “array” edge [10] [10] as can be seen in Figure 7.

To generate the edge binary map, vertical edge and horizontal edge detection is carried out. The algorithms for vertical edge and horizontal edge detection are as follows:

/ * Vertical edge detection 7

A1 = ABS (Ptrlmage [O] - Ptrlmage [1]);

A2 = ABS (Ptrlmage [O] - Ptrlmage [-1];

If (((A1> Th) && (A2> Th)) || (A1> 5 * Th / 2) || (A2)> 5 * Th / 2)) Edge [m] [n] = 1; / * edge 7

O {/ * Horizontal edge detection 7

ΑΊ = ABS (Ptrlmage [O] - PtrlmagefwidthJ);

A'2 = ABS (Ptrlmage [O] - Ptrlmagef-widthJ;

If (((A'1> Th) && (A'2> Th)) || (A'1> 5 * Th / 2) || (A'2)> 5 * Th / 2))

Edge [m] [n] = 1; / * edge 7

For vertical edge detection, an absolute value (Al) of the difference between the pixel gradient operation results (Ptrlmage [0]) is calculated in which it is determined whether the pixel constitutes an edge of the block, and is calculated the right pixel (Ptrlmage [I]) of the pixel (Ptrlmage [0]). An absolute value (A2) of the

ES 2 328 439 B2 differentiates between the gradient operation results of the Ptrlmage [0] and the left pixel (Ptrlmage [-1]) of the pixel (Ptrlmage [0]). Then, it is determined whether the pixel constitutes an edge according to the logical values ​​obtained after a comparison of the absolute values ​​A1 and A2 is made with a predetermined threshold value Th, and then the above procedure is carried out on all the pixels of the block. Vertical edge detection is carried out according to a logical formula of (A1> Th) && (A2> Th) II (A1> 5 * Th / 2) II (A2)> 5 * Th / 2). If the logical formula is true, the pixel is determined as a vertical border. Otherwise, the pixel is determined not to be a vertical border.

The horizontal edge detection is carried out on the same principle as the vertical edge detection. First, an absolute value (A'1) of the difference between the pixel gradient operation results (Ptrlmage [0]) is calculated in which it is determined whether the pixel forms an edge of the block, and the pixel is calculated lower (Ptrlmage [width]) of the pixel (Ptrlmage [0]). An absolute value (A'2) of the difference between the gradient operation results of the Ptrlmage [0] and the upper pixel (Ptrlmage [-width]) of the pixel (Ptrlmage [0] is then calculated. if the pixel constitutes an edge according to the logical values ​​obtained, after making a comparison of the absolute values ​​A'1 and A'2 with a predetermined threshold value Th and then the above procedure is carried out on all the pixels of the block. The horizontal edge detection is carried out according to a logical formula of (A'1> Th) && (A'2> Th) II (A'1> 5 * Th / 2) II (A'2)> 5 * Th / 2). If the logical formula is true, the pixel is determined as a horizontal border. Otherwise, the pixel is determined not to be a horizontal border. In this case, "&&" represents the logical AND expression, and "II" represents the logical OR expression.

Filtering is then carried out by applying a predetermined filter window to the generated edge binary map. Filtering can be done using a general procedure that applies a filter window of predetermined size. However, here filtering is not carried out if the central pixel of the filter window constitutes an edge, while filtering is performed if the central pixel does not constitute an edge. The filter window can be general. In this case, a 4-connectivity filter window is used that presents five pixels arranged in a cross shape, one placed in the center, as illustrated in Figure 7. In said Figure 7, X represents an edge pixel and the Zones other than zones containing X represent pixels that do not constitute the border.

Also, if the filter window has no edge pixels, ordinary filtering is performed, while weighted filtering is performed if there are edge pixels. An example of the weighted filtering is shown in Figure 7. In Figure 7, "<<" represents a shift to the left, and ">>" represents a shift to the right.

The invention can be carried out on a general digital computer that executes a program from a computer-readable medium, including but not limited to storage media, such as magnetic storage media (e.g. ROM, floppy disks, disks hard, etc.), optically readable media (for example, CD-ROMs, DVDs, etc.) and carrier waves (for example, Internet transmissions). Accordingly, the present invention may be realized as a computer-capable means comprising a computer-readable program code unit for adaptive signal filtering, comprising the computer-readable program coding means arranged in the medium. capable of being processed by computer: computer-readable program encoding means for a computer to perform block information generation to reduce the effect of blocks and hum information to reduce hum noise, from determined pixel coefficients of the zones of upper and left limit of the data block when a frame obtained by decomposing image data into bitstream for inverse quantization is an intraframe; computer-readable program encoding means for a computer to set the previous frame buzz and block information corresponding to a motion vector as current frame buzz and block information if the frame is an interframe, and set the hum information at "1" representing image data that needs to be filtered if there is a residual signal from the current block inversely quantized; and computer-readable program coding means for a computer to adaptively filter image data passing through block-based inverse quantization and inverse discrete cosine transform based on information that is generated about blocks and the buzz, for example. From the present description of the invention, functional program, code and code segments that can be used for the application of the present invention will be apparent to a skilled computer programmer.

Industrial applicability

As described above, the present invention can remove block noise and hum noise from a restored image from a compressed, block-based image, thereby enhancing the restored image from compression.