The Research Of Image Deblocking Artifacts Algorithm Based On Sparse Representation And Image Decomposition

Transform coding based on the block discrete cosine transform(BDCT) has been widely used in image and video compression standards. However, at low bit rates, decoded images will appear visible blocking artifacts along the block boundaries, affecting visual effect seriously. In order to reducing blocking artifacts, this paper focuses on image deblocking algorithm considering image sparse representation and image decomposition. The main contents are as follows:First, utilizing overcomplete sparse representation of image, and combining with alternating projection algorithm, the paper proposes three-times-iterative deblocking artifacts algorithm. The first two iteration use HIO algorithm for projection, and the third iteration uses the ER algorithm for projection. Experiment results show that the reconstruction effect of the deblocking algorithm based on alternating projection than the deblocking algorithm using one-time sparse representation is better.Secondly, considering the fact that the first order total variation of image is easy to produce staircase effect, then, we use the first total variation and second order total variation to reduce blocking artifacts. Considering the fact that a lot of denoising results of image denoising algorithms offen lose useful information of noise image, we propose a deblocking algorithm with two steps for improving the quality of reconstructed image, using the first order total variation for denoising in the first step, extracting useful information via sparse representation of residual image in the second step. Experiment results show that the deblocking algorithm of using the second order total variation of image and the deblocking algorithm with two steps have better reconstrcted effect than the algorithm only using the first order total variation of image.Finally, basing on image decomposition, and utilizing total variation and wavelet tree structure prior of image, the deblocking artifacts algorithm based on cartoon-texture image decomposition by total variation and dual tree complex wavelet transform using bivariate threshold rule is proposed. Experiment results show that this method can reduce blocking artifacts and keep much more marginal structures and detail information of image, it improves the quality of reconstructed image effectively and the reconstruction results are in good visual effect.