基于TVGA正则化的室内运动模糊图像恢复  

Regularized Motion Blurred Image Restoration Based on Topkis-Veinott Gradient Approach in Indoors

作　　者：胡春华[1] 马旭东[1] 戴先中[1] 

机构地区：[1]东南大学自动化学院,南京210096

出　　处：《中国图象图形学报》2007年第9期1531-1536,共6页Journal of Image and Graphics

基　　金：国家高技术研究发展"863"计划资助项目(2004AA420110)

摘　　要：针对室内环境下相机曝光时间长,被摄目标相对相机运动会产生图像模糊的问题,以室内服务移动机器人为研究背景,提出了一种基于Topkis-Veinott梯度法(TVGA)正则化运动的模糊图像恢复方法。此算法首先采用方向导数法估计出运动模糊方向,同时将图像运动模糊方向旋转至水平轴;然后采用自相关函数平均法确定运动模糊长度,并算出运动模糊点扩展函数(PSF);最后采用改进的TVGA最优化正则参数,进而恢复原始图像。与经典的Wiener法和两种正则化恢复方法进行的比较结果表明,用TVGA法正则化恢复的图像效果较好,不仅较接近原始图像,且易于实现。The relative motion between the camera and object in indoor would lead to blur image of the object. In this paper, Topkis-Veinott gradient approach（TVGA） is introduced to restore the motion-blurred images. Firstly the direction derivation algorithm is proposed to identify the motion-blurred direction from blurred image and then the motion-blurred direction can be adjusted to horizontal axis by image rotation. Secondly the average of autocorrelation functions of the blurred image derivative lines in horizontal direction is calculated, and then the blurred length of point spread function （PSF） is obtained by calculating the horizontal distance from the center to the lowest point. And then the modified TVGA is employed to optimize the regularization parameter. The algorithm has been compared with traditional. Wiener algorithm and other two image restoration methods based on regularization respectively. And the experimental results indicate that the TVGA has better restoration results and is easier to realize.

关 键 词：运动模糊 图像恢复 点扩展函数 正则化 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]