Image Enhancement Based on Noise Estimation

A novel multiparameter system for accurate enhancement of noisy images is presented. The proposed system includes cascaded blocks performing directional smoothing and sharpening whose behavior is finely controlled by a set of parameters. Since the most important parameter settings depend upon the noise, the standard deviation of the Gaussian noise is estimated by an algorithm that is based on piecewise linear filtering. Thus, simple functions (or look-up tables) are adopted to choose satisfactory parameter values for the overall enhancement system. Computer simulations and experiments with images acquired by digital cameras show that the proposed approach yields effective results