Block Artifact Reduction by Optimization Utilizing Subband Decomposition

The block discrete cosine transform (BDCT) is extensively used for compression of image and videos. One of its disadvantages, however, is the blocking artifacts at the block boundaries at low bit-rates. Most previous techniques for removing block artifacts utilize low-pass ltering or projection-on-convex-set (POCS) concept that gives an iterative restoration. These techniques produce unnecessary blurring in the image and require high computational power. In this paper, we introduce a new method for removing blocking eeects using an optimization technique and localized ltering through subband decomposition. We use transform domain ltering (TDF) technique in order to reduce the computational complexity. The subband decomposition allows localized ltering around the block boundaries where block artifacts would occur. The DCT coeecient quantization error correction utilizes Kuhn-Tucker theorem which is a Lagrangian method with inequalty constraints .