Remote Sensing Image Denoising Base-classified Mixture Modeling of Wavelet-domain

摘要

an algorithm about remote sensing image denoising base-classified mixture modeling of wavelet-domain is proposed in this paper. Firstly, remote sensing image is dealt with by the wavelet transform; being the fact that the difference of the wavelet coefficients between the image and noise, wavelet coefficients are classified into two categories: significant coefficients and insignificant ones by an adaptive threshold classification.then the differnent statistical models are proposed according to different categories of wavelet coefficients. The significant coefficients are denoised by bivariate shrinkage model with interscale depency, and the insignificant coefficients are denoised by zero-mean Gaussian model with high local correlation. Finally, a simplified “Cyclespinning” method is used to suppress the artifacts such as gibbs phenomena that may exist in the denoised images. Experimental results showed that this algorithm is effective both in removing noise and in reserving the image edge.