Iterative blind image restoration using local constraints

A new method of incorporating local image constraints into blind image restoration is proposed. The local mean and variance of the degraded image are used to obtain an initial estimate of the pixel intensity bounds. As the restoration proceeds, the bounds are updated from the current image estimate. The iterative-bound algorithm shows an improvement over the use of fixed bounds taken from the blurred image, in which case underestimation of the variance occurs at edges and textures. Simulations are presented for both the fixed- and iterative-bound implementations.