基于张量的块循环迭代去噪算法  

Tensor-based Block-cyclic Iterative Denoising Algorithm

作　　者：魏家濠 Wei Jiahao(School of Mathematical Sciences,Guizhou Normal University,Guiyang,China)

机构地区：[1]贵州师范大学数学科学学院,贵州贵阳

出　　处：《科学技术创新》2023年第10期22-25,共4页Scientific and Technological Innovation

摘　　要：彩色图像去噪在有效性方面具有挑战性。近年来,通过图像块集群利用图像相似性和变换域稀疏性在彩色图像去噪方面取得优异的性能。但是许多相关方法不能充分利用图像的非局部自相似性进行图像去噪,造成去噪后的图像过于光滑和细节丢失,因而去噪效果不好。本文为了保留图像信息和增强图像去噪的效果,我们提出基于t-SVD的迭代算法Rt-SVD,利用张量保留原始彩色图像的结构信息,用迭代将原始图像的信息加以考虑,提升搜索相似块的精确度,更加充分利用图像的非局部自相似性,达到最大程度保留原始图像细节的可能性,同时在真实数据集上Rt-SVD算法证明了有效性和鲁棒性。Color image denoising is challenging in terms of effectiveness.In recent years,excellent performance has been achieved in color image denoising by using image block clustering to leverage image similarity and transform domain sparsity.However,many related methods cannot fully utilize the non-local self-similarity of images to denoise them,resulting in overly smooth and detail-loss images and poor denoising effects.In order to preserve image information and enhance denoising effects,we propose the iterative algorithm Rt-SVD based on t-SVD,which uses tensors to preserve the structural information of the original color image and iteratively considers the information of the original image to improve the accuracy of searching for similar blocks and more fully utilize the non-local self-similarity of images,thus maximizing the possibility of preserving the original image details.Moreover,the effectiveness and robustness of the Rt-SVD algorithm are demonstrated on real data sets.

关 键 词：彩色图像去噪 非局部自相似性 t-SVD 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]