Image data post-processing method for reducing quantization effect, apparatus therefor

An edge detection method for deringing filtering to reduce ringing noise, the method comprising:

• detecting a horizontal edge of an image data;
• and
• detecting a vertical edge of the image data,
• wherein when a pixel within a block is pixel[m][n], a pixel to the right of the pixel[m][n] is pixel[m][n+1], a pixel to the left of the pixel[m][n] is pixel[m][n−1], the difference between the pixel[m][n] and the pixel[m][n+1] is A1, and a difference between the pixel[m][n] and the pixels[m][n+1] is A2, and a quantization factor of H.263 is QP, horizontal edge detection is performed by a logical formula ((A1>QP) and (A2>QP)) or (A1>2QP) or (A2>2QP) wherein the pixel[m][n] is determined to be an edge if the logical formula is satisfied, and
• when the pixel above the pixel[m][n] is pixel[m+1][n], the pixel below the pixel[m][n] is pixel[m−1][n], the difference between the pixel[m][n] and the pixel[m+1][n] is A′1, the difference between the pixel[m][n] and the pixel[m−1][n] is A′2, and the quantization factor of H.263 is QP, vertical edge detection is performed by a logical formula ((A′1>QP) and (A′2>QP)) or (A′1>2QP) or (A′2>2QP) wherein the pixel[m][n] is determined to be an edge if the logical formula is satisfied.

A deringing filtering method for reducing ringing noise, comprising:

• detecting horizontal and vertical edges of image data;
• and
• performing 2-D signal adaptive filtering on a block requiring reduction of ringing noise,
• wherein when a pixel within the block is pixel[m][n], a pixel to the right of the pixel[m][n] is pixel[m][n+1], a pixel to the left of the pixel[m][n] is pixel[m][n−1], a difference between the pixel[m][n] and the pixel[m][n+1] is A1, and a difference between the pixel[m][n] and the pixels[m][n+1] is A2, and a quantization factor of H.263 is QP, a horizontal edge detection is performed by a logical formula ((A1>QP) and (A2>QP)) or (A1>2QP) or (A2>2QP) wherein the pixel[m][n] is determined to be an edge if the logical formula is satisfied, and
• when a pixel above the pixel[m][n] is pixel[m+1][n], a pixel below the pixel[m][n] is pixel[m−1][n], a difference between the pixel[m][n] and the pixel[m+1][n] is A′1, a difference between the pixel[m][n] and the pixel[m−1][n] is A′2, the vertical edge detection is performed by a logical formula ((A′1>QP) and (A′2>QP)) or (A′1>2QP) or (A′2>2QP) wherein the pixel[m][n] is determined to be an edge if the logical formula is satisfied, and
• signal adaptive filtering is performed by applying a 4-connected filter window to the block, wherein filtering is not performed if a central pixel of the filter window is an edge pixel, and weighted filtering is performed if the central pixel of the filter window is a non-edge pixel.

A method of detecting corner outliers occurring at a corner of a cross point where four blocks meet when image data compressed based on a block are decoded, when four pixels around the cross point are pixels A, B, C and D, value[0] is A, value[1] is B, value[2] is C, value[3] is D, (A+B+C+D+2)/4 is Average, A10 and A20 are pixels adjacent to the pixel A in a block to which the pixel A belongs, A30 is a pixel diagonal to the pixel A, B10 and B20 are pixels adjacent to the pixel B in a block to which the pixel B belongs, B30 is a pixel diagonal to the pixel B, C10 and C20 are pixels adjacent to the pixel C in a block to which the pixel C belongs, C30 is a pixel diagonal to the pixel C, D10 and D20 are pixels adjacent to the pixel D in a block to which the pixel D belongs, and D30 is a pixel diagonal to the pixel D, the method comprising:

• (a) comparing a difference between the value[0] and the Average with the quantization factor (QP) of H.263, and designating the pixel A as the corner outlier candidate pixel if the difference is greater than the QP;
• (b) performing the step (a) on the value[1], value[2] and value[3] to designate a corresponding pixel as a corner outlier candidate pixel;
• and
• (c) designating the candidate pixel as a corner outlier pixel if the corner outlier candidate pixel is only one, and designating the candidate pixel having the greatest difference from (A3+B3+C3+D3+2)/4 as a corner outlier pixel if there are two or more corner outlier candidate pixels.

A method of compensating for detected corner outliers occurring at a corner of a cross point where four blocks meet when image data compressed based on a block are decoded, assuming that four pixels around the cross point are pixels A, B, C and D, value[0] is A, value[1] is B, value[2] is C, value[3] is D, (A+B+C+D+2)/4 is Average, A10 and A20 are pixels adjacent to the pixel A in a block to which the pixel A belongs, A30 is a pixel diagonal to the pixel A, B10 and B20 are pixels adjacent to the pixel B in a block to which the pixel B belongs, B30 is a pixel diagonal to the pixel B, C10 and C20 are pixels adjacent to the pixel C in a block to which the pixel C belongs, C30 is a pixel diagonal to the pixel C, D10 and D20 are pixels adjacent to the pixel D in a block to which the pixel D belongs, and D30 is a pixel diagonal to of the pixel D, the method comprising the steps of:

• compensating for the pixel A by (4A+B+C+2D+4)/8, the pixel A1 by (A′+3A1+2)/4, the pixel A2 by (A′+3A2+2)/4, if the corner outlier pixel is A and the difference between the pixels A and A3 is less than 3QP/2;
• compensating for the pixel B by (4B+C+D+2A+4)/8, the pixel B1 by (B′+3B1+2)/4, the pixel B2 by (B′+BA2+2)/4, if the corner pixel is B and the difference between the pixels B and B3 is less than 3QP/2;
• compensating for the pixel C by (4C+D+A+2B+4)/8, the pixel C1 by (C′+3C1+2)/4, the pixel C2 by (C′+3C2+2)/4, if the corner pixel is C and the difference between the pixels C and C3 is less than 3QP/2;
• and
• compensating for the pixel D by (4D+A+B+2C+4)/8, the pixel D1 by (D′+3D1+2)/4, the pixel D2 by (D′+3D2+2)/4, if the corner pixel is D and the difference between the pixels D and D3 is less than 3QP/2.

An image data post-processing method for reducing corner outliers occurring at a corner of a cross point where four blocks meet when image data compressed based on a block are decoded, the method comprising:

• (a) detecting corner outliers from the block of inverse-quantized image data;
• and
• (b) compensating for the detected corner outliers,
• wherein when four pixels around the cross point are pixels A, B, C and D, value[0] is A, value[1] is B, value[2] is C, value[3] is D, (A+B+C+D+2)/4 is Average, A10 and A20 are pixels adjacent to the pixel A in a block to which the pixel A belongs, A30 is a pixel diagonal to the pixel A, B10 and B20 are pixels adjacent to the pixel B in a block to which the pixel B belongs, B30 is a pixel diagonal to the pixel B, C10 and C20 are pixels adjacent to the pixel C in the block to which the pixel C belongs, C30 is a pixel diagonal to the pixel C, D10 and D20 are pixels adjacent to the pixel D in a block to which the pixel D belongs, and D30 is a pixel diagonal to the pixel D, the step (a) comprises the sub-steps of:
• (a1) comparing a difference between the value[0] and the Average with the quantization factor (QP) of H.263, and designating the pixel A as the corner outlier candidate pixel if the difference is greater than the QP;
• (a2) performing the step (a1) on the value[1], value[2] and value[3] to designate a corresponding pixel as a corner outlier candidate pixel;
• and
• (a3) designating the candidate pixel as a corner outlier pixel if the corner outlier candidate pixel is only one, and designating the candidate pixel having the greatest difference from (A3+B3+C3+D3+2)/4 as a corner outlier pixel if there are two or more corner outlier candidate pixels, and
• wherein the step (b) is performed by compensating for the pixel A by (4A+B+C+2D+4)/8, the pixel A1 by (A′+3A1+2)/4, the pixel A2 by (A′+3A2+2)/4, if the corner outlier pixel is A and the difference between the pixels A and A3 is less than 3QP/2;
• compensating for the pixel B by (4B+C+D+2A+4)/8, the pixel B1 by (B′+3B1+2)/4, the pixel B2 by (B′+BA2+2)/4, if the corner outlier pixel is B and the difference between the pixels B and B3 is less than 3QP/2;
• compensating for the pixel C by (4C+D+A+2B+4)/8, the pixel C1 as C1′ by (C′+3C1+2)/4, the pixel C2 by (C′+3C2+2)/4, if the corner outlier pixel is C and the difference between the pixels C and C3 is less than 3QP/2;
• and compensating for the pixel D by (4D+A+B+2C+4)/8, the pixel D1 as D1′ by (D1+3D1+2)/4, the pixel D2 by (D′+3D2+2)/4, if the corner outlier pixel is D and the difference between the pixels D and D3 is less than 3QP/2.

An image data post-processing method for reducing quantization effect induced when image data compressed based on a block is decoded, the method comprising:

• (a) detecting a semaphore representing whether post-processing is required, using distribution of inverse quantization coefficients of inverse-quantized image data and a motion vector representing a difference between blocks of previous video object plane (VOP) and blocks of a current VOP;
• and
• (b) detecting a corner outlier pixel of the inverse-quantized image data block, when four pixels around the cross point are pixels A, B, C and D, value[0] is A, value[1] is B, value[2] is C, value[3] is D, (A+B+C+D+2)/4 is Average, A10 and A20 are pixels adjacent to the pixel A in a block to which the pixel A belongs, A30 is a pixel diagonal to the pixel A, B10 and B20 are pixels adjacent to the pixel B in a block to which the pixel B belongs, B30 is a pixel diagonal to the pixel B, C10 and C20 are pixels adjacent to the pixel C in a block to which the pixel C belongs, C30 is a pixel diagonal to the pixel C, D10 and D20 are pixels adjacent to the pixel D in a block to which the pixel D belongs, and D30 is a pixel diagonal to the pixel D, by:
• (b1) comparing a difference between the value[0] and the Average with the quantization factor (QP) of H.263, and designating the pixel A as the corner outlier candidate pixel if the difference is greater than the QP;
• and
• (b2) performing the step (b1) on the value[1], value[2] and value[3] to count the corresponding pixel as a corner outlier candidate pixel;
• (c) filtering the decoded image data corresponding to the semaphore, if it is determined by checking the detected semaphore that the post-process is required;
• and
• (d) designating the pixel as a corner outlier pixel if the corner outlier candidate pixel is only one and then compensating for the pixel by a predetermined method if the difference between the pixel and a pixel diagonal to the pixel is less than 3QP/2, and designating the candidate pixel having a greatest difference from (A3+B3+C3+D3+2)/4 as a corner outlier pixel if there are two or more corner outlier candidate pixels and then compensating for the detected candidate pixel by the predetermined method.

An image data post-processing apparatus for reducing quantization effect induced when image data compressed based on a block is decoded, the apparatus comprising:

• a semaphore detector for detecting a semaphore representing whether post-processing is required, using distribution of inverse quantization coefficients of inverse-quantized image data and a motion vector representing the difference between blocks of previous video object plane (VOP) and blocks of a current VOP wherein the semaphore includes a blocking semaphore representing whether reducing blocking artifacts near block boundaries is required, and a ringing semaphore representing whether or reducing ringing noise near image edges is required;
• a deblocking filter for checking the blocking semaphore detected by the semaphore detector and performing deblocking filtering on the decoded image data;
• a corner outlier compensator for detecting a corner outlier from the data passed through the deblocking filtering and compensating for the detected corner outlier;
• and
• a deringing filter for checking the ringing semaphore detected by the semaphore detector and performing deringing filtering on the corner outlier compensated data.

A computer readable medium having embodied thereon a computer program for image data post-processing capable of reducing quantization effect induced when image data compressed based on a block is decoded,

• wherein the image data post-processing method comprises:
• detecting a semaphore representing whether post-processing is required on an intra video object plane (VOP) in an intra-VOP mode and an inter-VOP in an inter-VOP mode, by investigating a distribution of inverse quantization coefficients of inverse-quantized image data and calculating a motion vector representing a difference between blocks of previous VOP and blocks of current VOP;
• and
• filtering the decoded image data corresponding to the semaphore, if it is determined by checking the detected semaphore that post-processing is required.

A computer readable medium having embodied thereon a computer program for image data post-processing capable of reducing corner outliers occurring at the corner of a cross point where four blocks meet when image data compressed based on a block are decoded,

• wherein the image data post-processing method comprises:
• (a) detecting corner outliers from the block of inverse-quantized image data;
• and
• (b) compensating for the detected corner outliers,
• wherein when four pixels around the cross point are pixels A, B, C and D, value[0] is A, value[1] is B, value[2] is C, value[3] is D, (A+B+C+D+2)/4 is Average, A10 and A20 are pixels adjacent to the pixel A in a block to which the pixel A belongs, A30 is a pixel diagonal to the pixel A, B10 and B20 are pixels adjacent to the pixel B in a block to which the pixel B belongs, B30 is a pixel diagonal to the pixel B, C10 and C20 are pixels adjacent to the pixel C in the block to which the pixel C belongs, C30 is a pixel diagonal to the pixel C, D10 and D20 are pixels adjacent to the pixel D in a block to which the pixel D belongs, and D30 is a pixel diagonal to the pixel D, the step (a) comprises the sub-steps of:
• (a1) comparing a difference between the value[0] and the Average with the quantization factor (QP) of H.263, and designating the pixel A as the corner outlier candidate pixel if the difference is greater than the QP;
• (a2) performing the step (a1) on the value[1], value[2] and value[3] to designate a corresponding pixel as a corner outlier candidate pixel;
• and
• (a3) designating the candidate pixel as a corner outlier pixel if the corner outlier candidate pixel is only one, and designating the candidate pixel having the greatest difference from (A3+B3+C3+D3+2)/4 as a corner outlier pixel if there are two or more corner outlier candidate pixels, and
• wherein the step (b) is performed by compensating for the pixel A by (4A+B+C+2D+4)/8, the pixel A1 by (A′+3A1+2)/4, the pixel A2 by (A′+3A2+2)/4, if the corner outlier pixel is A and the difference between the pixels A and A3 is less than 3QP/2;
• compensating for the pixel B by (4B+C+D+2A+4)/8, the pixel B1 by (B′+3B1+2)/4, the pixel B2 by (B′+BA2+2)/4, if the corner outlier pixel is B and the difference between the pixels B and B3 is less than 3QP/2;
• compensating for the pixel C by (4C+D+A+2B+4)/8, the pixel C1 as C1′ by (C1+3C1+2)/4, the pixel C2 by (C′+3C2+2)/4, if the corner outlier pixel is C and the difference between the pixels C and C3 is less than 3QP/2;
• and compensating for the pixel D by (4D+A+B+2C+4)/8, the pixel D1 as D1′ by (D′+3D1+2)/4, the pixel D2 by (D′+3D2+2)/4, if the corner outlier pixel is D and the difference between the pixels D and D3 is less than 3QP/2.