Modifying weber fraction law to postprocessing and edge detection applications

A postprocessing scheme is proposed to enhance compressed images. The main objective is to obtain improvements that are pertinent to the properties of human visual system. The proposed scheme implements Weber fraction (also called contrast sensitivity) to enhance the appearance of the current block by incorporating information from adjacent blocks. The ratio DeltaI/I is found between the mean of a line of pixels in the current block and two points, each resides on the boundary of an adjacent block, that are the continuation of the chosen line. To avoid biasing toward low intensity values and to preserve the symmetry of the sensitivity curve, I was replaced by the maximum of the actual mean value and the corresponding value of the negative image or simply max(mean, 255-mean). If DeltaI/I is less than a threshold, the chosen line is replaced with a one fitting the original data and the two boundary points. Although PSNR improvement is <0.3 dB, the resultant image is visually more pleasing as will be demonstrated experimentally. The algorithm can be easily modified to perform as an edge detection scheme by finding DeltaI/I between any pixel and its 8 neighbours. The maximum is then taken. A new histogram thresholding is then applied to discriminate the edge pixels. Experimental results indicate a superior capability of the proposed scheme to detect edges of objects that are close in intensity to their background. Some comparisons with Sobel operator are also demonstrated.