Nonlinear adaptive restoration of images with multiplicative noise

Assuming a pointwise digital model of multiplicative noise, restoration is defined as optimal mean-squared-error estimation from an observed image using known statistical properties of the degrading noise field. An approach to restoration that uses locally adaptive multiplicative operators with the property of multiplicative superposition is formulated. Adaptive point and two-point algorithms that are based on Taylor series expansions of optimal formulas and the normalization of data by local multiplicative sample means are developed. Image preprocessing results on multiplicatively noisy images are demonstrated and discussed