earticle

영어

In this paper, we propose an improved method for the removal of additive Gussian white noise from PET-CT images. Different from the traditional sparse representation based denoising methods, our method is composed of two distinctively steps such as the preliminary denoise and the detail compensation. By constructing a sparse representation model, denoising is formulated as an optimization problem that can be solved on an over-complete dictionary. The proposed method effectively trains this dictionary by using K-SVD algorithm with atom replace model. Then the preliminary denoised image is reconstructed through improved OMP algorithm with the fidelity factor of SSIM (Structural Similarity). The detail compensation image is obtained by using the difference between the noisy image and the preliminary de-noised image, and the improved OMP algorithm is employed again to get the denoised detail compensation image. Finally, the final denoised image is reconstructed by adding the denoised detail compensation image to the preliminary denoised image. Experiment results have shown that the proposed method is better than some other denoising methods in terms of PSNR and SSIM.