quantizer defined by the so-called quantization matrix. This quantization operation introduces an error in the re-It is well known that blocking effects constitute one of the main drawbacks of discrete-cosine-transform-based lossy codconstructed image, which is almost imperceptible for moding schemes. In this paper, we first propose a distortion measure erate compression ratios. A problem arises, however, for for blocking artifacts in images. The visibility of blocking error higher compression ratios: the algorithms generate the is computed based on several characteristics of the observer's well-known blocking artifacts (also called grid noise) which visual perception, and an image quality factor is derived based severely degrade the perceived quality of the images. on the artifact visibilities. We then develop a novel approach Blocking effect has been found to be very objectionable for removing blocking effects in transform-coded images. The to the viewer, especially due to the perceptual characteristechnique is based on a nonlinear, spatially variant adaptive tics of this particular distortion [4]. Hence, the quality of smoothing operation which exploits the visibility measure of a reconstructed picture coded by DCT would be greatly blocking artifacts present in the decoded image. It introduces improved if the block distortion were suppressed. no modification on actual image compression standards and Several methods already have been studied to reduce allows one to reduce the blockiness without significantly blurblocking effect [5-10]. Some iterative methods are based ring the image. Experimental results obtained at low bit rate on the theory of projections onto convex sets (POCS), but show that both objective and subjective quality of the recona great number of iterations can be needed to efficiently structed images are greatly improved.