JP2926638B2 - Loop filter and loop filtering method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to data filtering, and more particularly, to a loop filter and a loop filtering method for reducing blocking effects and ringing noise.

[0002]

2. Description of the Related Art Generally, MPEG of International Standards Organization (ISO)
H. of ITU (International Telecommunication Union)
Most image coding standards, including 263, employ block-based motion estimation and block discrete cosine transform (DCT) processing. The block-based coding leads to a blocking effect and ringing noise, as is well known, especially when the image is highly compressed. Typical blocking effects include lattice noise in homogeneous regions having relatively similar pixel values between adjacent pixels, and staircase noise in which image edges appear stepwise along the edge of the image. Can be Further, the ringing noise is a typical Gibbs phenomenon generated when the coefficients of the discrete cosine transform (DCT) are truncated by quantization in order to highly compress the image.

When the compressed data is restored and displayed on a screen, the trace of the grid noise appears on the edge of the block, so that the edge between the blocks becomes noticeable. In addition, since the edge of the image of the staircase noise still appears in a stepped manner, the edge of the image becomes uneven. Then, the ringing noise causes a problem that a plurality of objects appear to overlap at predetermined intervals.

Several methods have been proposed to reduce the blocking effect and ringing noise that occur during such block-based encoding. First, H.261
Then, the blocking effect is reduced by using a simple 3 × 3 low-pass filter as a loop filter. Also, simple edge loop filters have been proposed to reduce blocking effects and mosquito noise. The edge loop filter uses a method of linearizing two pixel values at the boundary of a block and changing two pixel values into the linearized value. However, although the edge loop filter can reduce the blocking effect, it has a disadvantage that ringing noise is not reduced.
A non-linear filter using a binary index has been proposed to reduce ringing noise. However, there is a disadvantage that the problem related to the blocking effect cannot be solved.

[0005]

SUMMARY OF THE INVENTION The present invention has been devised to solve the above problems, and at the same time, greatly reduces the blocking effect and ringing noise generated when encoding based on the block. SUMMARY OF THE INVENTION It is an object of the present invention to provide a loop filter and a loop filtering method for reducing the blocking effect and ringing noise of a high-compression coding system that performs simple low-pass filtering.

[0006]

According to the present invention, there is provided a loop filtering method for reducing a blocking effect and ringing noise of image data according to the present invention. Generating binary edge map information by comparing a result value calculated using the degree operator with a preset threshold value; and setting a predetermined value for the generated binary edge map information. Determining whether the binary edge map information included in the filter window includes edge information by applying a filter window of a size, and determining that the edge information does not include edge information in the determining step And filtering a pixel value belonging to the filter window for each pixel using a preset first weight to generate a new pixel value. A generation step, when it is determined that that contain edge information in said determining step, the filter c
When the pixel located at the center of the window is not edge information
A new pixel value is generated by filtering a pixel value belonging to the filter window for each pixel using a preset second weight value, and a pixel located at the center of the filter window is edge information. If it is is characterized in that it comprises a second generating step without filtering.

The gradient operator in the step of generating the binary edge map information has a size of 1 × 2, and its weight is (1, −−).
1) and a vertical inclination operator having a size of 2 × 1 and a weight value of (1, −1). The generating of the binary edge map information may include receiving data in units of a predetermined block size for the image, and using a one-dimensional gradient operator for each pixel of the input image block. Calculating the gradient, and setting the first pixel to which the pixel for which the gradient has been calculated belongs to a predetermined area around the block boundary.
A binary edge map information is generated by comparing with a threshold value, and a binary edge map is generated by comparing with a preset second threshold value if the pixel whose slope is calculated does not belong to a predetermined area around a block boundary. Generating a binary edge for generating information.

Further, a loop filter according to the present invention for achieving another object of the present invention comprises: an image storage unit for temporarily storing image data; and an image storage unit for storing image data in blocks of a predetermined size from the image storage unit. Acceptance, a gradient calculation unit that performs one-dimensional gradient calculation in the horizontal and vertical directions using a one-dimensional gradient operator to search for edge pixels, and the pixel that has performed the horizontal and vertical one-dimensional calculation is in the boundary area An area judging unit for judging whether the image data belongs to a comprehensive area which is not the boundary area or not, and when the image data calculated by the horizontal or vertical one-dimensional gradient operator is judged as a boundary area by the area judging unit. A boundary area binary edge information generation unit that generates edge information when the threshold value is larger than a predetermined threshold value and generates non-edge information that is not an edge when the threshold value is lower than the predetermined threshold value. When the image data calculated by the horizontal or vertical one-dimensional gradient operator is determined to be a comprehensive area by the area determining unit, if the image data is larger than a predetermined threshold value, the image data is generated as edge information, and the predetermined threshold value The comprehensive area binary edge information generating unit that generates non-edge information that is not an edge if the following is obtained, and values calculated by a vertical gradient operator and a horizontal gradient operator for one pixel are determined by the area determination. OR operation unit for ORing information generated through the unit and the boundary area binary edge information generation unit or the comprehensive area binary edge information generation unit to generate edge map information, and supplied from the OR operation unit. 2
A filter determining unit that stores binary edge map information and classifies input image data into an edge region including at least one edge information and a homogeneous region not including edge information according to the binary edge map; An average filter that performs a predetermined average filtering on a filter window pixel determined to be a homogeneous region by the unit,
A weighting filter for performing a predetermined weighting filtering on the filter window pixel determined to be an edge region by the filter determination unit.

[0009]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, the present invention will be described in more detail with reference to the accompanying drawings. FIG. 1 is a block diagram showing the configuration of an encoder to which a loop filter according to the present invention can be applied as a preferred embodiment, and shows an encoder related to H.263. In the encoder of FIG. 1, blocks other than the loop filter 140 according to the present invention are known to those skilled in the art of high-compression encoding systems including motion-compensation-prediction encoders. The operation will be briefly described.

In FIG. 1, an input video signal is subtracted by a subtractor 10.
0 is input to the first switch 105 and the predictive encoder 145. The subtractor 100 subtracts the motion-estimated error signal output from the predictive encoder 145 from the input video signal, and the output signal of the subtractor 100 is applied to a first switch 105. The first switch 105 selects one of the input video signal and the output signal of the subtractor 105 according to the control of the encoding controller 120, and the selected signal is output to the orthogonal transformer 110.
Is applied to The encoding controller 120 controls the first switch 100 to apply the input video signal to the orthogonal transformer 110 in the case of the intra encoding mode, and to apply the output signal of the subtractor 100 to the orthogonal transformer 110 in the case of the inter encoding mode. The voltage is applied to the converter 110. The signal passing through the orthogonal transformer 110 and the quantizer 115 is applied to an inverse quantizer 125 and a video multiplexing encoder (not shown). The inverse quantizer 115 inversely quantizes the applied signal, and the inversely quantized signal is applied to an inverse orthogonal transformer.
Applied to 130. The signal subjected to the inverse orthogonal transformation by the inverse orthogonal transformer 130 and the signal subjected to the prediction encoding by the prediction encoder 145 are applied to a loop filter 140. The loop filter 140 removes blocking effects and ringing noise. The predictive encoder 145 receives the signal filtered by the loop filter 140 and the input video signal, performs motion estimation, and outputs a motion-estimated error signal. Further, the prediction encoder 145 outputs the motion vector (v) obtained by the motion compensation prediction encoding to the video multiplex encoder.

The output signal of the predictive encoder 145 is a subtractor
100 and the second switch 150. Second switch 15
0 is controlled by the coding controller 120. In the case of the intra coding mode, the signal processed by the predictive coder 145 is blocked from being supplied to the adder 135, and the inter coding mode is controlled. In this case, the input signal is supplied to the adder 135. The coding controller 120 controls the operation of the apparatus shown in FIG. 1, and includes an intra / inter mode information flag (p) indicating whether to perform intra-frame coding or inter-frame coding, A signal (qz) indicating the conversion characteristic designation information, a flag (t) enabling discrimination between transmission and non-transmission, and a signal (q) indicating the level of the conversion coefficient are generated.

FIG. 2 is a block diagram for explaining the loop filter 140 of FIG. The loop filter of FIG. 2 includes an image storage unit 200, a binary edge map information generation unit 210, and a signal adaptive filter unit 250. The image storage unit 200 temporarily stores the image data including the blocking effect and ringing noise applied from the inverse orthogonal transformer 130 of FIG. 1 and the signal output from the predictive encoder 145. The binary edge map information generation unit 210 receives an image signal from the image storage unit 200 in units of 16 × 16 macroblocks composed of 16 × 16 pixels, and receives a one-dimensional gradient in units of 8 × 8 blocks. Binary edge map information is generated by using an operator, and the slope degree calculating section 212, the area determining section 214, the boundary area binary edge information generating section 216, and the comprehensive area binary edge information generating section 218 are provided.
And a logical sum operation unit 220. The signal adaptive filter unit 250 is a block for filtering the 16 × 16 size binary edge map information generated from the binary edge map information generating unit 210, and a filter determining unit 252.
, An average filter 254 and a weighting filter 256. Here, the size of the macroblock and the block is a value for one embodiment of the present invention, and it is apparent to those skilled in the art that the present invention is not limited to the above size.

The inclination operator 212 is provided in the image storage unit 20.
Accepts image data in units of 16 × 16 size macroblocks consisting of 0 to 16 × 16 pixels, and performs horizontal and vertical one-dimensional gradient operations using a one-dimensional gradient operator to search for edge pixels. Do. The one-dimensional inclination operator used here has a value of (1, -1) as a weight value, and has a 1 × 2 horizontal one-dimensional inclination used when performing the inclination operation in the horizontal direction. A degree operator and a vertical one-dimensional slope operator having a size of 2 × 1 used when performing a slope operation in the vertical direction.

The area determining unit 214 determines whether the pixel subjected to the horizontal or vertical one-dimensional inclination calculation belongs to a boundary area or belongs to a comprehensive area other than the boundary area. The division between the boundary area and the inclusive area is to remove the blocking effect by taking different threshold values according to the area when generating the binary edge information.

When the image data calculated by the horizontal or vertical one-dimensional gradient operator is determined to be a boundary area by the area determining unit 214, the boundary area binary edge information generating unit 216 determines a predetermined threshold value. If it is larger, it is generated as edge information, and if it is less than the predetermined threshold value, it is generated as non-edge information that is not an edge. Here, it is preferable that the predetermined critical value is set to 200.

When the image data calculated by the horizontal or vertical one-dimensional gradient operator is determined to be a comprehensive area by the area determining unit 214, the comprehensive area binary edge information generating unit 218 has a predetermined threshold value. If it is larger, it is generated as edge information, and if it is less than the predetermined threshold value, it is generated as non-edge information that is not an edge. Here, it is preferable that the predetermined critical value is set to 10.

The OR operation unit 220 calculates the value calculated by the vertical one-dimensional inclination operator and the value calculated by the horizontal one-dimensional inclination operator for one pixel.
214 and the boundary area binary edge information generation unit 216 (or
If the information is generated as separate information through the comprehensive area binary edge information generation unit 218), it is ORed and generated as edge information.

FIG. 3 shows the binary edge map information generator 21.
0 shows a binary edge map generated at 0 and a low-pass filter used in the signal adaptive filter unit 250.
The filter determination unit 252 stores the binary edge map information provided from the logical sum operation unit 220. Input image data is 2
It is classified into an edge area and a homogeneous area according to the advanced edge map. In the present invention, the filter determination unit 252 is used for the above-described classification. The averaging filter 254 and the weighting filter 256 according to the embodiment of the present invention use a 3 × 3 size filter window. Therefore, the filter window used in the filter determining unit 252 also has a size of 3 × 3. The filter determining unit 252 determines whether the region on the binary edge map where the filter window is located is an edge region based on an edge value included in a preset size, that is, a 3 × 3 size filter window. It is determined whether the area is a homogeneous area. If it is determined that the area is the homogeneous area, the filter determining unit 252 outputs the position data for the center point in the filter window used at the time of the determination to the average filter 254. If it is determined that the area is an edge area, the filter determining unit 252 outputs to the weighting filter 256 the binary edge map information in the filter window used in the determination and the position data for the center point. Here, the center point means a point at which the value of the pixel located at that point is replaced with a new value by filtering.

FIGS. 4A to 4C show two-dimensional 3 × 3
FIG. 4A shows a filter window for a 3 × 3 filter. FIG. 4B shows weight values for the 3 × 3 average filter, and FIG. 4C shows weight values for the 3 × 3 weight filter. In the filter window of FIG. 4A, the point where the filter weight index value is “5” is the center point of the filter window.

Average filter 254 and weight filter 25
Reference numeral 6 denotes a kind of two-dimensional low-pass filter. The operation of the filters 254 and 256 will be described in more detail as follows. When the location data for the center point is entered,
The average filter 254 reads the pixel values necessary for calculating the filtered pixel value of the center point from the image storage unit 200. Next, the average filter 254 calculates the filtered pixel value using the read pixel value and the weight value shown in FIG. 4B. The calculated filtered pixel value is used as a modified pixel value for the center point. The weighting filter 256 is a filter determining unit 252
Performs a filtering operation on the basis of the binary edge map information provided from and the position data for the center point. The operation of the weighting filter 256 will be described in more detail as follows. If the center point "5" belongs to an edge point, the weighting filter 256 does not perform a filtering operation on the center point. If the edge point (or a plurality of edge points) has a position in the 3 × 3 filter window excluding the center point, the weighting filter 256 performs a filtering operation using the weighting values shown in FIG. . When an arbitrary edge point is located at 2 and 6, 6 and 8, 4 and 8, or 2 and 4 in FIG. 4 (A), the weight values of the edge point and the adjacent points outside the edge point are all "0". . Average filter 254
The weighted filter 256 outputs image data that has been subjected to signal adaptive filtering.

On the other hand, the operation of the present invention based on the above configuration will be described as follows. FIG. 5 is a flowchart showing the operation of the present invention. First, in the intra mode, the image data including the blocking effect and the ringing noise applied from the image storage unit 200 from the inverse orthogonal transformer 130 of FIG. 1 is converted into a 16 × 16 size macro consisting of 16 × 16 pixels. A block is received and a one-dimensional gradient operation is performed using a horizontal or vertical one-dimensional gradient operator in blocks of 8 × 8 size (500 steps). More specifically, the input image data is received in units of 18 × 18 macroblocks composed of 18 × 18 pixels. This is because the target processing unit is a macroblock of 16 × 16 size, but pixels of an adjacent macroblock are required to process a pixel on the boundary.

Next, it is determined through the area determining unit 214 whether the pixel for which the inclination has been calculated belongs to the boundary area or the comprehensive area (step 505). If it belongs to the boundary area, the horizontal or vertical boundary area threshold values (Tbh, Tbv) preset through the boundary area binary edge information generation unit 216 are set.
(510 step). When the horizontal inclination operator is used, if the value is larger than the Tbh, information indicating an edge "
1 "is generated, otherwise non-edge information"
Similarly, when the vertical gradient operator is used, information “1” indicating an edge is generated if the value is larger than Tbv, and information “0” that is not an edge is generated otherwise. (Step 515) Here, Tbv and Tbh are set to 200.

On the other hand, if it belongs to the comprehensive area, it is compared with a preset horizontal or vertical comprehensive area threshold value (Tgh, Tgv) through the comprehensive area binary edge information generator 218 (step 520). Then, when the horizontal inclination operator is used, information “1” indicating an edge is generated if the horizontal inclination operator is larger than the Tgh, and information “0” which is not an edge is generated otherwise. Similarly, when the vertical gradient operator is used, if it is larger than the Tgv, information "1" indicating an edge is generated, and otherwise information "0" which is not an edge is generated (step 525). here,
Tgv and Tgh are set to 10.

If the binary edge information generated by using the horizontal and vertical gradient operators is different as described above,
The logical sum is calculated through the logical sum operation unit 220 (step 530).
Edge information “1” is generated. By performing the above-described process for each pixel, binary edge map information is generated in units of 16 × 16 macroblocks (step 535).

The generated binary edge map information is 3
Filtered through a filter having a × 3 size. The filtering process will be described in more detail. First, a filtering area is set for image data of 8 × 8 size with a filter size of 3 × 3 for each pixel, and the filter determination unit 252 determines whether or not there is a pixel indicating edge information among the pixels in the filtering area. (Step 540). As a result, if there is edge information, it is checked whether the center pixel of the filter indicates edge information (step 545). if,
If the edge information is indicated, the pixel value of the original input image data is used as it is without filtering (step 550). If not, weighted filtering is performed using the weighted filter 256 (step 555). On the other hand, if there is no pixel indicating edge information in the filtering area in operation 540, average filtering is performed using the average filter 254 (operation 560). As a result, the filtered image data forms loop filtered image data in units of 16 × 16 macroblocks,
By repeating this, loop filtering is performed on the frame image.

[0026]

According to the present invention, since the blocking effect and the ringing noise are removed from the image decompressed and restored based on the blocks, the quality of the decompressed image can be improved. It is obvious that the present invention is not limited to the above-described embodiment, and that many modifications can be made by those skilled in the art within the technical idea to which the present invention belongs. In addition, although the above-described embodiment has been described with reference to H.263, it can be similarly used in the case of VM (Verification Model) 3.1 of MPEG-4.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of an encoder to which a loop filter according to the present invention can be applied as a preferred embodiment;

FIG. 2 is a block diagram showing a configuration of a loop filter shown in FIG.

FIG. 3 shows a binary edge map generated by a binary edge map information generation unit and a low-pass filter used in a signal adaptive filter unit.

FIG. 4A shows a filter window for a two-dimensional 3 × 3 filter, and FIGS. 4B and 4C show two-dimensional 3 × 3 filters.
Shown are the weights for the x3 mean filter.

FIG. 5 is a flowchart illustrating a loop filtering method according to the present invention.

[Explanation of symbols]

 200 Image storage section 210 Binary edge information map generation section 212 Slope degree calculation section 214 Area determination section 216 Boundary area binary edge information generation section 218 Comprehensive area binary edge information generation section 220 OR operation section 250 Signal adaptive filter section 252 Filter determination unit 254 Average filter 256 Weighted filter

──────────────────────────────────────────────────続 き Continued on the front page (58) Fields surveyed (Int. Cl. ⁶ , DB name) G06T 5/00-5/50 H04N 7/24-7/68 H04N 1/41-1/419

Claims (9)

(57) [Claims] 1. A loop filtering method for reducing a block effect and ringing noise of image data, comprising: a) a result value calculated using a predetermined one-dimensional gradient operator for each pixel of an image and a preset value; Generating binary edge map information by comparing the generated threshold value with the threshold value; and b) applying a filter window of a predetermined size to the generated binary edge map information. Determining whether the binary edge map information belonging to the doe includes edge information; and c) if it is determined that the edge information does not include edge information in the determining step, the pixel belongs to the filter window for each pixel. A first generation step of generating a new pixel value by filtering the pixel value using a preset first weight value; and d) including edge information in the determination step. It is determined to have Ru
Sometimes an image located in the center of the filter window
If the element is not edge information, a new pixel value is generated by performing filtering using a preset second weight for a pixel value belonging to the filter window for each pixel,
Wherein when the pixel located at the center of the filter window is an edge information, the loop filtering method characterized in that it comprises a second generating step without filtering. 2. A loop filtering method according to claim 1, wherein said filter window has a filter window having a size of 3 × 3. 3. The gradient operator at the stage of generating the binary edge map information has a size of 1 × 2, and a horizontal gradient operator whose weight is (1, −1); The loop filtering method according to claim 1, wherein the vertical filtering operator has a size of 1 and its weight is (1, -1). 4. The step of generating binary edge map information includes the steps of: a-1) accepting data in units of a predetermined block size for an image; and a-2) receiving each pixel of the input image block. A-3) calculating a gradient using a one-dimensional gradient operator, and a-3) a first threshold value that has been preset as the pixel for which the gradient has been calculated belongs to a predetermined area around a block boundary. And generating the binary edge map information by comparing the calculated pixel with a predetermined second threshold value if the calculated pixel does not belong to a predetermined area around the block boundary. 2. The loop filtering method according to claim 1, further comprising the steps of: 5. The step of generating binary edge map information includes the steps of: a-1) accepting data in units of a predetermined block size for an image; and a-2) receiving each pixel of the input image block. A-3) calculating a gradient using a one-dimensional gradient operator, and a-3) a first threshold value that has been preset as the pixel for which the gradient has been calculated belongs to a predetermined area around a block boundary. And generating the binary edge map information by comparing the calculated pixel with a predetermined second threshold value if the calculated pixel does not belong to a predetermined area around the block boundary. 4. The loop filtering method according to claim 3, comprising the steps of: 6. The method as claimed in claim 4 , wherein in the step of generating the binary edge, the first threshold value is larger than the second threshold value. 7. The one-dimensional horizontal slope calculation and the one-dimensional vertical slope calculation are respectively performed on each pixel of the image block in the slope calculation step, and the one-dimensional horizontal slope calculation is performed in the binary edge generation step. 5. The loop filtering method according to claim 4, wherein a binary value is generated for the degree calculation and the one-dimensional vertical slope calculation, and a binary edge map information is generated using a result value obtained by performing a logical sum. . 8. The block is 16 × 16 in size,
5. The method according to claim 4, wherein the loop filtering is performed in units of blocks each having a size of 8 * 8. 9. An image storage unit for temporarily storing image data, and receiving image data in blocks of a predetermined size from the image storage unit, and using a one-dimensional gradient operator to search for edge pixels. And a gradient calculating unit that performs a one-dimensional gradient calculation in the vertical direction; and a region determination unit that determines whether the pixel that has performed the horizontal and vertical one-dimensional calculations belongs to a boundary region or belongs to a comprehensive region that is not the boundary region. When the image data calculated by the horizontal or vertical one-dimensional gradient operator is determined to be a boundary region by the region determining unit, if the image data is larger than a first threshold value, the image data is generated as edge information; A boundary area binary edge information generating unit that generates non-edge information that is not an edge if the value is equal to or less than a first threshold value; and an image data calculated by the horizontal or vertical one-dimensional gradient operator. When the area is determined to be a comprehensive area by the area determination unit, edge information is generated when the area is larger than a second threshold, and non-edge information that is not an edge is generated when the area is less than the second threshold. A comprehensive region binary edge information generating unit, and a value calculated by a vertical gradient operator and a horizontal gradient operator for one pixel is used as the region determining unit and the boundary region binary edge information generating unit or the comprehensive region. An OR operation unit for ORing information generated through a binary edge information generation unit to generate edge map information; and storing binary edge map information supplied from the OR operation unit. A filter determining unit for classifying into an edge region including at least one edge information and a homogeneous region not including edge information according to a binary edge map; An average filter for performing a predetermined average filtering on the filter window pixels determined to be an area, and a weighting filter for performing a predetermined weighted filtering to the filter window pixels determined to be the edge area by the filter determination unit. And a loop filter.