Blind and adaptive image restoration in the framework of a multiscale Gabor representation
Gabriel Cristóbal, Rafael Navarro · 2002
We present a blind method for space-variant image restoration, which can be applied to remove degradations due to either additive noise or blurring, but not both simultaneously. Restoration is performed using a multiscale Gabor (1946) representation, which consists of a pyramid of 4/spl times/4 Gaborian channels in frequency and orientation plus a zero-frequency residual. The definition of a generic undegraded image prototype in the Gabor domain, by averaging over many high quality (undegraded) images, permits one to carry out a blind recursive image restoration. The method has been tested with images degraded by computer simulated additive noise (Brownian fractal) and two examples of real blurring, obtaining good results in all cases.>