Abstract

In many applications that necessitate image recovery the imaging system can be modeled as a convolution of the original image with an ill-posed operator and then adding noise. We assume that the M×M original image is blurred by a space-invariant convolution operator and independent identically distributed (i.i.d.) zero mean additive noise described by the model where H is a M²×M² Toeplitz matrix modeling the space-invariant blur, the vectors g, f, and n are M²×1 lexicographic orders of the degraded observed image, the original image and the noise, respectively.

© 1992 Optical Society of America