一种基于局部Radon变换运动模糊图像参数估计算法  

Algorithm of Motion Blur Parameter Estimation Based on Local Radon Transform

作　　者：廖永忠[1,2] 蔡自兴[1] 何湘华[2] 

机构地区：[1]中南大学信息科学与工程学院,长沙410081 [2]湖南第一师范学院信息科学与工程系,长沙410205

出　　处：《小型微型计算机系统》2014年第1期133-136,共4页Journal of Chinese Computer Systems

基　　金：湖南省教育厅科研项目(09C230)资助

摘　　要：由于相机的运动会使获取的图像变得模糊,而图像模糊过程的本质是一个卷积过程.恢复这类模糊图像的关键是准确的估计运动模糊核函数,即对运动方向和模糊长度这两个参数的估计.提出一种基于局部Radon变换的方法来解决运动模糊核函数估计的问题,通过对模糊图像的频谱进行局部Radon变换,以此来构造一个目标函数,然后求解该目标函数的L1范数最优解来估计运动模糊的方向,在模糊长度估计上,提出一种改进的启发式算法,该算法能实现不同模糊程度下模糊长度的估计.实验结果表明,本文的算法在低信噪比和小模糊长度的情况下仍能准确的估计模糊核参数.Camera movement during exposure will result in the image blur,its essence is a convolution process. The key problems of restoring this kind of images is how to estimate the blur kernel function,namely the accurate estimation of motion direction and blur length from an observed image is performed. In this paper, a novel algorithm is presented for motion blur parameters estimation by processing the local Radon transform of the image spectrum, this method can estimate the direction of motion by solving the L1 norm optimal solution of the objective function, an improved heuristic algorithm based on degree of the blur is used to estimate the blur length. Experimental results show that the algorithm in low SNR and small blur length of the case still correct estimate of the blur parameters.

关 键 词：运动模糊 点扩散函数 RADON变换 参数估计 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]