An Image Denoising Method for Noises Based on Mixed Gaussian Distribution in HSV Color Space

摘要

Traditional image noise model is created for color images in RGB color space. Differing with the traditional model, noise model is created in HSV space towards origins of noise generation in this paper. It is supposed that noise images based on mixed Gaussian distribution with zero mean value on three dimensions, including hue, saturation and value, are quite similar to actual situations. For images are characterized by simple background, EM algorithm is applied to estimation of noise parameters and related parameters are applied to wavelet threshold denoising in this paper. Results of the experiment show that both estimation and denoising are satisfactory.