An image restoration algorithm based on image fusion

An image restoration algorithm based on two-step iterative shrinkage (TwIST) and multiply filter banks is proposed, which combines the multi-scale, multi-resolution, multi-direction, spatial anisotropy of contourlet transform, and the multi-scale, multi-resolution and spatial isotropy of wavelet transform. First, the blurred image is separately deblurred by various wavelet basis and contourlet basis based on TwIST, so as to select the better deblurred images. Second, these deblurred images are fused into one clean image based on the fused measure. The weighted average method is applied to the fused measure. The power is decided by the residual between the blurred image and the deblurred image. The deviations of the local variances of the residual image are exploited to evaluate the local quality of the deblurred image in pixel domain. The experimental results show that the proposed algorithm achieves improvement on improved signal-noise ratio (ISNR). © 2012 Praise Worthy Prize S.r.l.