Perceptual Denoising of Color Images

摘要

Denoising of color images is a trade-off between sharpness of an image and perceived noise. We formulate a novel optimization problem that can maximize sharpness of an image while limiting the perceived noise under a model of visibility of additive random noise. We derive a closed-form expression for an optimal two-dimensional finite impulse response filter, show its uniqueness and existence, and present simulation results for black and white as well as color images. Simulation results show remarkable reduction in perceptibility of noise, while preserving sharpness. The computational burden required for the optimal filter is reduced by a new ad hoc filter which is simple but has near optimal performance.