【作者】 王文颖；

【导师】 洪景新；

【作者基本信息】 厦门大学 ， 信号与信息处理， 2009， 硕士

【摘要】 在成像过程中,相机与物体之间的相对旋转运动会导致得到的图像中某些有用信息被掩盖。旋转运动模糊因其模糊的空间变化的特性,近年来成为数字图像处理领域的主要研究课题之一。该类型的运动模糊是红外制导系统中普遍存在的问题,会影响目标的识别和跟踪,因此,对旋转运动模糊图像的复原有着非常重要的理论和实际应用价值。由于实际获得的旋转运动模糊图像的模糊参数不可控制,因此本文在进行算法研究时,首先根据模糊图像的退化模型生成了一系列仿真旋转运动模糊图形,以此作为前期研究素材,在证实了算法的正确性后,再进行实际验证。在对模糊图像进行复原之前,本研究先从水平运动模糊尺度的确定方法展开,提出了一种利用自相关来鉴别旋转模糊角度的方法,并利用仿真图像进行了算法正确性的验证。旋转模糊角度的鉴别是确定图像退化点扩展函数的关键,在此基础上,本文讨论了维纳滤波、有约束最小二乘方滤波和LR滤波的性能,并最终选择了维纳滤波作为理论基础,提出了一种旋转运动模糊图像的复原算法,利用bresenham画圆算法的思想将空间变化模糊转变为空间不变模糊,利用维纳滤波实现一维的去卷积,再变换回原空间,达到图像复原的目的。该算法应用于仿真图像,论证了算法的可行性,然后针对实际图像旋转圆心不可知的特性,分别选取不同的像素点作为旋转中心,对实际得到的模糊图像进行处理,可以从中提取字迹图案等有用信息,达到了本论文的预期目的。更多还原

【Abstract】 In imaging process,the relative rotation motion will lead to some useful message be covered.The Characteristic of rotation motion blur is space-variant. Therefore the restoration of rotation motion blurred image becomes one of main problem in image processing research.This kind of motion blur will impact object identification and tracking in Infrared Guidance System.The restoration of rotation motion blurred image has important value of theory and engineering application.A series of simulation rotation motion blurred images are formed according to rotation motion blurred degradation model.These simulation blurred images are research objects in Preliminary Stage.When the correctness of algorithm is verified, the practical obtained blurred images are processed to confirm.Before the restoration of blurred images,the rotation angle is identified at first.Three widely used methods for the identification of rectilinear motion blurred extent,Fourier transform,error-parameter curve,and autocorrelation,are introduced in rotation motion blurred image analysis.Then,the autocorrelation is proved to be an effective method by processing the simulation images.The performances of three filtering methods are analyzed,Wiener Filtering, constrained least-squares Filtering,and Lucy-Richardson Filtering.And then,a restoration method of rotation motion blurred images based on wiener filtering is proposed.In this method,the space-variant blurs are transformed to space-invariant blurs by using the Bresenham circle-producing algorithm.Deconvolutions are practiced by using wiener filtering.Then,the space-invariant are transformed back to space-variant in order to realize image restoration.The feasibility of this method is proved by processing the simulation blurred images.And then,focusing on the uncertain of the rotation circle center,the practical obtained blurred images are restored with different circle centers.The word and design in restored blurred images can be extracting.The research purpose is realized basically.更多还原