Abstract:

Due to its high phase transition performance and low computational complexity, approximate message passing (AMP) becomes a powerful reconstruction scheme in large-scale applications such as image reconstruction. How to employ the image structured sparsity priors is a key subject for the study of AMP image reconstruction. The cartoon-texture model is introduced into the AMP image reconstruction. According to the different characters of the cartoon and texture component, a hierarchical filtering operator is designed for AMP image reconstruction based on dualtree complex wavelet transform and total variation. Effects of the iterations on the structural characteristics of the to be denoised images and the performance of the filtering operators are analyzed. An AMP image reconstruction algorithm based on the cartoon-texture model and staging filtering is proposed. Experimental results show that this algorithm can preserve more contour and texture information, and improve the quality of the reconstructed images.