视频监控图像的运动模糊方向估计  

Motion blurred direction estimation for video monitor image

作　　者：吴海滨[1] 刘超[1] 于晓洋[1] 孙晓明[1] 胡丽娟[1] 

机构地区：[1]哈尔滨理工大学测控技术与仪器黑龙江省高校重点实验室,黑龙江哈尔滨150080

出　　处：《液晶与显示》2014年第4期580-585,共6页Chinese Journal of Liquid Crystals and Displays

基　　金：哈尔滨市科技创新人才研究专项资金(No.2013RFQXJ090);黑龙江省博士后科研启动项目(No.LBHQ12069);哈尔滨理工大学青年拔尖创新人才培养计划;国家自然科学基金(No.61005035)

摘　　要：模糊图像复原是安防监控领域的一项重要技术,为提高模糊方向的估计准确度从而提高复原清晰度,对模糊图像的频谱分析、方向提取以及干扰抑制等方法进行研究.通过频谱图细化分块消除由边缘截断效应导致的十字亮线干扰.通过Radon变换归一化降低该算法固有的对角线干扰.最后,修正频谱图中条纹倾斜角度与模糊方向的偏差,以提高任意高宽比图像模糊方向的估计准确度.实验结果表明,避免了模糊方向估计曲线在0°、90°、45°（135°）方向出现干扰峰值,在非对角线方向能较准确地估计模糊方向,误差约4～6°.能有效抑制十字亮线干扰和对角线干扰,且适用于任意高宽比图像.Motion target detection plays an important role in object tracking and traffic monitoring fields.For improving the estimation accuracy of motion direction,and thereby increasing restoration resolution,some research of blurred image spectrum analysis,blurred direction detection and interfe rence suppression were conducted.The interference of bright cross in the spectrum caused by edge truncation effect was eliminated through partitioning spectrum image.The diagonal interference was reduced by normalized Radon transform.For improving estimation accuracy of blurred direction of any ration image,the factor aspect ratio was used to modify the relation between tilt angle of the dark stripe and motion direction angle.The experiment demonstrates that this algorithm can well avoid interference peak appeared in 0°,90°,45°（135°） direction of blurred direction estimation curve,and well estimate blurred direction in non diagonal with about 4～ 6 °error.Overall,this algorithm can suppress the interferences of bright cross and diagonal effectively,and applies to any aspect ratio.

关 键 词：视频监控 运动模糊 模糊方向估计 归一化Radon变换 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]