Infrared image denoising via sparse representation over redundant dictionary

Infrared images are often contaminated by much noise, thus it is significant to denoise the infrared image. An effective denoising method is presented in this paper. The infrared images are assumed with strong zero-mean white and homogeneous Gaussian adaptive noise. Focus on denoising image with high noise level, firstly, the image is denoised via sparse representation over an adaptive redundant dictionary. The dictionary is trained by applying K-means Singular Value Decomposition (K-SVD) algorithm on the down-scaled noisy image. Secondly, a double-scale denoising is added to improve the denoised results. The experimental results indicate that this method could obtain a better performance when noise level is high.