Methods and systems for detection and estimation of compression noise

Abstract

Systems and methods for estimation of compression noise in an image or in a video sequence are provided. Difference values in a direction are computed, and one or more statistical feature values for the difference values are computed in association with a plurality of offsets. At least one feature value is computed using the plurality of statistical feature values, and a compression-noise estimate may be determined in accordance with a criterion based on a comparison of the at least on feature value and an associated threshold and an additive noise estimate.

What is claimed is: 
     
         1 . A method for estimating noise in an image, the method comprising:
 receiving image data corresponding to a region of pixel values defined by a height and a width;   determining a number of offsets for the region of pixel values, wherein an offset defines a sub-set of pixel locations within the region;   for each offset, computing the standard deviation of respective directional difference values computed for each pixel location included in a sub-set;   computing the maximum of the computed standard deviations;   computing the mean of the computed standard deviation; and   computing a noise estimate in the image based on a function of the computed maximum of the standard deviations and the computed mean of the standard deviations.   
     
     
         2 . The method of  claim 1 , wherein a directional difference for a pixel location is computed as the difference between a pixel value at a current pixel location and a pixel value at an adjacent pixel location. 
     
     
         3 . The method of  claim 2 , wherein an adjacent pixel location is a horizontal adjacent pixel location. 
     
     
         4 . The method of  claim 3 , wherein an offset defines a sub-set of pixel locations as pixel locations included in every i-th column of the region. 
     
     
         5 . The method of  claim 3 , wherein an adjacent pixel location is a vertical adjacent pixel location. 
     
     
         6 . The method of  claim 5 , wherein an offset defines a sub-set of pixel locations as pixel locations included in every i-th row of the region. 
     
     
         7 . The method of  claim 1 , wherein the function of the maximum of the computed standard deviations and the mean of the computed standard deviations includes a weighted average function. 
     
     
         8 . A method for estimating noise in an image, the method comprising:
 receiving image data corresponding to a region of pixel values defined by a height and a width;   determining a number of offsets for the region of pixel values, wherein an offset defines a sub-set of pixel locations within the region;   for each offset, computing the standard deviation of respective directional difference values computed for each pixel location included in a sub-set, wherein a directional difference for a pixel location is computed as the difference between a pixel value at a current pixel location and a pixel value at an adjacent pixel location;   computing one or more statistics of the computed standard deviations; and   computing a noise estimate in the image based on a function of the one or more computed statistics.   
     
     
         9 . The method of  claim 8 , wherein the one or more statistics includes one or more of: a mean, a maximum, a median, a skew, and a kurtosis. 
     
     
         10 . The method of  claim 8 , wherein an adjacent pixel location is a horizontal adjacent pixel location. 
     
     
         11 . The method of  claim 10 , wherein an offset defines a sub-set of pixel locations as pixel locations included in every i-th column of the region. 
     
     
         12 . The method of  claim 8 , wherein an adjacent pixel location is a vertical adjacent pixel location. 
     
     
         13 . The method of  claim 10 , wherein an offset defines a sub-set of pixel locations as pixel locations included in every i-th row of the region. 
     
     
         14 . An apparatus for estimating noise in an image, the apparatus comprising:
 an image receiver for receiving image data corresponding to a region of pixel values defined by a height and a width;   an offset determination unit for determining a number of offsets for the region of pixel values, wherein an offset defines a sub-set of pixel locations within the region;   a directional difference calculator for computing the standard deviation of respective directional difference values computed for each pixel location included in a sub-set for each offset;   a statistic calculator for computing one or more statistics of the computed standard deviations; and   a noise estimator for computing a noise estimate in the image based on a function of the one or more computed statistics.   
     
     
         15 . The apparatus of  claim 14 , wherein the one or more statistics includes one or more of: a mean, a maximum, a median, a skew, and a kurtosis. 
     
     
         16 . The apparatus of  claim 14 , wherein a directional difference for a pixel location is computed as the difference between a pixel value at a current pixel location and a pixel value at an adjacent pixel location 
     
     
         17 . The apparatus of  claim 16 , wherein an adjacent pixel location is a horizontal adjacent pixel location. 
     
     
         18 . The apparatus of  claim 17 , wherein an offset defines a sub-set of pixel locations as pixel locations included in every i-th column of the region. 
     
     
         19 . The apparatus of  claim 16 , wherein an adjacent pixel location is a vertical adjacent pixel location. 
     
     
         20 . The apparatus of  claim 19 , wherein an offset defines a sub-set of pixel locations as pixel locations included in every i-th row of the region.