一种基于贝叶斯压缩感知的图像修复方法  

An Image Inpainting Method Based on Bayesian Compressive Sensing

作　　者：党宏社[1] 张娜[1] 

机构地区：[1]陕西科技大学电气与信息工程学院,西安710021

出　　处：《河南大学学报（自然科学版）》2014年第5期601-607,共7页Journal of Henan University:Natural Science

摘　　要：图像修复是利用图像中已知区域信息对破损区域进行信息填充,以弥补信息的损失.传统的修复方法依赖图像的结构来确定,使图像达到人眼主观可以接受的程度.基于贝叶斯压缩感知的图像修复方法首先对受损图像进行稀疏变换,利用贝叶斯压缩感知得到稀疏系数的后验分布函数,求得分布函数的均值和方差,将均值作为图像的稀疏系数的估计,方差作为噪声的估计.仿真结果验证了该方法可以提高图像的修复质量.Image inpainting is to use the known information in the image to fill the damaged region to make up for the lost information. The traditional inpainting methods are determined through the structure of image to make the image reach a subjective acceptable degree of the human eyes. The image inpainting method, on the basis of Bayesian compressive sensing, transforms the sparsity of the damaged image first, then gets the posterior distribution function of the sparse coefficient through Bayesian compressive sensing . At last, the mean and the variance of the distribution function are obtained. The mean can be used as an estimation of the sparse coefficient for the image, and the variance is an estimation of the noise. The emulation results proved that this method can improve the inpainting quality of images.

关 键 词：图像修复 贝叶斯压缩感知 后验分布 稀疏变换 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]