Adaptive postfiltering of transform coefficients for the reduction of blocking artifacts

This paper proposes a novel postprocessing technique for reducing blocking artifacts in low-hit-rate transform-coded images. The proposed approach wor ks in the transform domain to alleviate the accuracy loss of transform coef ficients, which is introduced by the quantization process. The masking effe ct in the human visual system (HVS) is considered, and an adaptive weightin g mechanism is then integrated into the postfiltering. In low-activity area s, since blocking artifacts appear to be perceptually more detectable, a la rge window is used to efficiently smooth out the artifacts. In order to pre serve image details, a small mask, as well as a large central weight, is em ployed for processing those high-activity blocks, where blocking artifacts are less noticeable due to the masking ability of local background. The qua ntization constraint is finally applied to the postfiltered coefficients. E xperimental results show that the proposed technique provides satisfactory performance as compared to other postfilters in both objective and subjecti ve image quality.