Deep Total Variation Denoising Network for Low-Dose CT Images
Research on denoising low-dose CT images was conducted,the issues of low computational performance and insufficient generalization in neural network denoising for artifact suppression were analyzed.An anisotropic total variation deep unfolding denois-ing network was adopted,with the new method incorporating the edge characteristics of adjacent voxels,the anisotropic TV regulari-zation term was introduced to preserve the structural information of images and avoid the edge blurring caused by isotropic TV.The Chambolle-Pock algorithm was employed to solve the mathematical model,suitable for deep unfolding into convolutional neural net-works.Additionally,the pixel attention mechanism was integrated for the network optimization to capture the important information of images.Through experimental tests on the Mayo 2016 dataset,this method has advantages over traditional methods and other ad-vanced network models in image denoising,showing a better performance in the indicators of PSNR,SSIM,and VIF.This method meets the requirements for high-quality reconstruction of low-dose CT images.
万方数据
维普
