一种基于块配准的运动模糊图像复原算法  

A MOTION-BLURRED IMAGE RESTORATION ALGORITHM BASED ON BLOCK REGISTRATION

作　　者：陶恂[1] 

机构地区：[1]上海电机学院电子信息学院,上海200240

出　　处：《计算机应用与软件》2011年第7期246-248,263,共4页Computer Applications and Software

摘　　要：提出一种基于图像块配准的运动模糊图像复原算法。首先对测试图像进行分块操作,图像块的大小根据经验值选取;其次,采用SIFT算法在模糊图像和清晰图像对之间进行块配准;遍历模糊图像中的每个图像块,分别计算得到各个块的质心坐标,然后利用质心法计算该图像块发生偏移的角度和距离;统计图像中所有块的偏移,估计出图像退化的点扩散函数,并对每个图像块进行复原校正。分别针对仿真图像和智能视频监控图像进行了实验,来验证算法的有效性。实验结果表明,该算法能够有效地复原智能交通领域的智能视频监控图像,并且具有较高的复原效率。A block registration based motion-blurred image restoration algorithm is presented.Firstly,a blocking operation is performed on test images,in which the image block sizes are determined by empirical values.Secondly,SIFT algorithm is applied to register block pairs between the blurry image and the clear one.Thirdly,each image block in the blurry image is traversed to obtain its centroid coordinates by its independent calculation;then the centroid method is applied to calculate the offset angle and distance of the image block.All block offsets are counted;the image degradation point spread function is estimated and applied to every image block for restoration.Experiments are carried out both on simulated images and on intelligent video surveillance images to validate the effectiveness of the algorithm.Experimental results show that the proposed algorithm can effectively restore intelligent traffic domain with comparatively high restoration efficiency.

关 键 词：图像复原 图像配准 运动模糊 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]