Spacial Deblocking Method Using Limited Edge Differences Only to Linearly Correct Blocking Artifact

Abstract

A de-blocking method smoothes pixels along a row or column that crosses a block boundary. Smoothing is performed to remove quantization or compression artifacts that appear on block edges when pixels in adjacent blocks are separately compressed. A maximum-allowed edge-pixel difference is generated from the quantization parameter QP. For each edge-crossing row or column, an edge difference is generated as half the difference between adjacent edge pixels in two blocks. This edge difference is compared to the maximum-allowed edge-pixel difference. When the edge difference is larger than the maximum-allowed edge-pixel difference, then the difference is limited to the maximum-allowed edge-pixel difference, since the pixel difference may be a real edge in the image. The limited or edge difference is then added or subtracted in decreasing amounts for several pixels in the row or column near the edge, smoothing the edge difference across several pixels, such as seven pixels.

1 . A filter comprising: 
 a maximum generator for generating a maximum-allowed edge-pixel difference as a function of a quantization parameter, the quantization parameter indicating a degree of compression of an image;    a difference generator, receiving pixel components from an edge pair of adjacent pixels in adjacent blocks, for generating an edge-pixel difference;    a limiter, receiving the maximum-allowed edge-pixel difference and the edge-pixel difference, for generating a difference term, the difference term being the edge-pixel difference when the edge-pixel difference is less than the maximum-allowed edge-pixel difference, but being the maximum-allowed edge-pixel difference when the edge-pixel difference is larger than the maximum-allowed edge-pixel difference;    a fractional difference generator, receiving the difference term from the limiter, for generating a plurality of fractional differences having a decreasing range of values that are fractions of the difference term; and    an applicator, receiving pixel components from a current row of pixels that includes the edge pair of adjacent pixels in the adjacent blocks, for adding or subtracting the fractional differences from the fractional difference generator to pixel components for pixels in the current row.    
   
   
       2 . The filter of  claim 1  wherein successive pixels successively farther from the edge pair of adjacent pixels have successively decreasing fractional differences added or subtracted,  
     whereby edge-pixel differences are smoothed over the current row of pixels.  
   
   
       3 . The filter of  claim 2  wherein the edge pair of adjacent pixels comprises a first current pixel in a current block and a first next pixel in a next block;  
     wherein the applicator further receives the difference term from the limiter, the applicator subtracting the difference term from the first current pixel, but adding a fractional difference to the first next pixel.  
   
   
       4 . The filter of  claim 3  wherein the applicator adds a fractional difference of three-fourths of the difference term to the first next pixel in the next block.  
   
   
       5 . The filter of  claim 4  wherein the current row of pixels comprises: 
 a second current pixel adjacent to the first current pixel in the current block;    a third current pixel adjacent to the second current pixel in the current block;    a fourth current pixel adjacent to the third current pixel in the current block;    a second next pixel adjacent to the first next pixel in the next block;    a third next pixel adjacent to the second next pixel in the next block.    
   
   
       6 . The filter of  claim 5  wherein the applicator adds a fractional difference of one-half of the difference term to the second next pixel in the next block; 
 wherein the applicator adds a fractional difference of one-fourth of the difference term to the third next pixel in the next block;    wherein the applicator subtracts a fractional difference of three-fourths of the difference term to the second current pixel in the current block;    wherein the applicator subtracts a fractional difference of one-half of the difference term to the third current pixel in the current block;    wherein the applicator subtracts a fractional difference of one-fourth of the difference term to the fourth current pixel in the current block.    
   
   
       7 . The filter of  claim 6  wherein the pixel components are luminance Y values of a YUV pixels.  
   
   
       8 . The filter of  claim 1  wherein the maximum generator generates the maximum-allowed edge-pixel difference as a linear function of a quantization parameter.  
   
   
       9 . The filter of  claim 8  wherein the maximum generator generates the maximum-allowed edge-pixel difference as one-half of the quantization parameter added to a constant.  
   
   
       10 . The filter of  claim 9  wherein the constant is four.  
   
   
       11 . The filter of  claim 1  further comprising: 
 a negator, receiving the maximum-allowed edge-pixel difference from the maximum generator, for generating a negative maximum-allowed edge-pixel difference;    wherein the limiter further comprises a negative limiter, receiving the negative maximum-allowed edge-pixel difference and the edge-pixel difference, for generating the difference term when the edge-pixel difference is negative, the difference term being the edge-pixel difference when the edge-pixel difference is between zero and the negative maximum-allowed edge-pixel difference, but being the negative maximum-allowed edge-pixel difference when the edge-pixel difference is more negative than the negative maximum-allowed edge-pixel difference.    
   
   
       12 . A computer-implemented method for smoothing blocking artifacts comprising: 
 receiving a series of pixels that cross a block boundary, the series comprising in sequence a first pixel, a second pixel, a third pixel, a fourth pixel, a fifth pixel, a sixth pixel, and a seventh pixel;    wherein the first, second, third, and fourth pixels are in a first block and the fifth, sixth, and seventh pixels are in a second block;    wherein the fourth pixel and the fifth pixel comprise edge pixels in the series of pixels;    generating a limit value from a quantization parameter that indicates a degree of compression of pixels in the first block;    generating a negated limit value from the limit value;    generating an edge difference value;    selecting the edge difference value as a limited difference when the edge difference value is positive and less than the limit value;    selecting the edge difference value as the limited difference when the edge difference value is negative and greater than the negated limit value;    selecting the limit value as the limited difference when the edge difference value is positive and greater than the limit value;    selecting the negated limit value as the limited difference when the edge difference value is negative and less than the negated limit value;    generating fractional differences from the limited difference; and    subtracting the limited difference from a pixel to smooth the pixel.    
   
   
       13 . The computer-implemented method of  claim 12  wherein the series of pixels comprise luminance Y values of YUV pixels.  
   
   
       14 . A computer-program product comprising: 
 a computer-usable medium having computer-readable program code means embodied therein for smoothing block artifacts produced during video compression, the computer-readable program code means in the computer-program product comprising:    input means for receiving a series of pixels de-compressed from a motion-picture-experts group (MPEG) compressed video stream, wherein a current series of pixels comprises at least five pixels;    maximum means for generating a maximum-allowed edge-pixel difference as a function of a quantization parameter, the quantization parameter indicating a degree of compression of an image;    difference generator means, receiving pixel components from an edge pair of adjacent pixels in adjacent blocks, for generating an edge-pixel difference;    limiter means, receiving the maximum-allowed edge-pixel difference and the edge-pixel difference, for generating a difference term, the difference term being the edge-pixel difference when the edge-pixel difference is between the maximum-allowed edge-pixel difference or a negation of the maximum-allowed edge-pixel difference, but being the maximum-allowed edge-pixel difference when the edge-pixel difference exceeds than the maximum-allowed edge-pixel difference or the negation of the maximum-allowed edge-pixel difference; and    fractional difference generator means, receiving the difference term from the limiter means, for generating a plurality of fractional differences having a decreasing range of values that are fractions of the difference term.    
   
   
       15 . The computer-program product of  claim 14  further comprising: 
 applicator means, receiving pixel components from the current series of pixels that includes the edge pair of adjacent pixels in the adjacent blocks, for adding or subtracting the fractional differences from the fractional difference generator means to pixel components for pixels in the current series;    wherein successive pixels successively farther from the edge pair of adjacent pixels have successively decreasing fractional differences added or subtracted,    whereby edge-pixel differences are smoothed over the current series of pixels.    
   
   
       16 . The computer-program product of  claim 15  wherein the edge pair of adjacent pixels comprises a first current pixel in a current block and a first next pixel in a next block; 
 wherein the applicator means further receives the difference term from the limiter means, the applicator means for subtracting the difference term from the first current pixel, and for adding a fractional difference to the first next pixel;    wherein the applicator means adds a fractional difference of three-fourths of the difference term to the first next pixel in the next block.    
   
   
       17 . The computer-program product of  claim 16  wherein the current series of pixels comprises: 
 a second current pixel adjacent to the first current pixel in the current block;    a third current pixel adjacent to the second current pixel in the current block;    a fourth current pixel adjacent to the third current pixel in the current block;    a second next pixel adjacent to the first next pixel in the next block;    a third next pixel adjacent to the second next pixel in the next block;    wherein the applicator means adds a fractional difference of one-half of the difference term to the second next pixel in the next block;    wherein the applicator means adds a fractional difference of one-fourth of the difference term to the third next pixel in the next block;    wherein the applicator means subtracts a fractional difference of three-fourths of the difference term to the second current pixel in the current block;    wherein the applicator means subtracts a fractional difference of one-half of the difference term to the third current pixel in the current block;    wherein the applicator means subtracts a fractional difference of one-fourth of the difference term to the fourth current pixel in the current block.    
   
   
       18 . The computer-program product of  claim 17  wherein the pixel components are luminance Y values of a YUV pixels.