Optimal distortion redistribution in block-based image coding using successive convex optimization

In the block-based image coding systems, the pixels at the block boundaries usually have larger distortions than the pixels at other positions. This biased distortion distribution not only generates unpleasant blocking artifact, but potentially decreases the intra-prediction performance and eventually diminishes the coding efficiency. Motivated by this observation, a novel distortion redistribution approach is proposed in this paper which aims to achieve the best rate-distortion performance. Our contributions are summarized in the following aspects: First, the inter-block dependency model (IBDM) is introduced as a quantitative measure of the inter-block dependency. Then, based on the proposed IBDM, the distortion redistribution problem is formulated as a rate minimization problem under a certain distortion constraint. In the end, successive convex optimization method is employed to tackle the optimization problem, and the final optimal distortion can be achieved efficiently. Experimental results demonstrate that compared with the original distortion distribution, significant gain can be obtained with the proposed approach.