Abstract

Surfacelet (ST) is a method that can be used in the sparse expression of 3-D signal which owns the characteristics like anisotropy, multi-direction and low redundancy. In this paper video signal is used as the input, non-local filter model Context is introduced to ST field which make the traditional 2-D model to 3-D. Along the subband direction, 3-D Context energy calculation model are built in each subbands which make the energy expression more precise According to the energy distribution, the decomposition coefficients are made into different classes by non-neighborhood and estimated thresholds are calculated for denoise process. From the results it can be seen that the proposed method is better at detail reservation and flicker restrain as well as blur, scratch, also the higher PSNR can be achieved.

Paper Details