A New Restoration Scheme for Defocus Blurred Image using Multiscale-Decomposition

The restoration of the photographically defocus blurred image depends primarily on the knowledge about the degradation process. There have been various methods to model the defocus blur. The effectiveness of the method for deriving the defocus blur model parameters determines the image restoration quality to some extent. In this work, assuming that the PSF can be modeled by a function with a single parameter, the value of this parameter is estimated from the low-frequency wavelet subband using multiscale-decomposition based on wavelet transform, and especially the reflection symmetry extending manner is adopted in the low-frequency wavelet subband to suppress the ringing ripples on the restored image. We reduce noise by using soft threshold denoising in the high-frequency wavelet subbands. The experimental results show that the proposed method is effect and can diminish computational complexity and speed up the restoration greatly.