模糊图像复原之研究

【作者】 姚芳兵；

【导师】 张铃；

【作者基本信息】 安徽大学 ， 计算机应用技术， 2004， 硕士

【摘要】 本文主要研究数字图像处理中常见的一类：模糊图像复原。图像复原，是指去除或减轻在获取数字图像过程中发生的图像质量下降(退化)；目标是对退化的图像进行处理，使它趋向于复原成没有退化的理想图像。商空间理论和遗传算法是模糊图像复原中可以借鉴的理论基础。 静态均匀散焦是在静态情况下的一种常见模糊，本文给出了判断模糊图像是否是均匀散焦的方法，通过对比实际的灰度图和理想能量图来判断，直观明了。在此基础上，应用了一种简明的复原算法：逆滤波器法，主要根据两函数卷积的傅立叶变换等于各个函数傅立叶变换的乘积这一原理而设计。同时，根据算法的特点提出了两种判断若干幅复原图清晰度的评价算法，从而得到最优复原图。 匀速直线运动模糊亦是动态情况下的一种典型模糊，因为变速的、非直线的运动在某些条件下可以看成是均匀的、直线运动的合成结果。同样运用了一种速度颇快的复原算法：差分法，把复杂的复原过程转变成几个简单的部分，每一部分只是累加求和。最后讨论了几类判断复原图清晰度的评价函数，力求能得到最清晰的复原图。 本文结合具体的模糊图验证算法和理论的正确性，从实验结果看，复原效果还算满意，不过仍有提高的空间。更多还原

【Abstract】 This paper mostly researches a kind in digital image processing: Restoration of Blurring Image. Image restoration is that it wipes off or lightens fall of image quality in the course of getting digital image, aims to deal with image of degradation and makes it to ideal image that has no degradation. Quotient-space and Genetic Algorithm are basic theory which can be use for reference at restoration of blurring image.Static uniform defocusing is a common kind of blurring in static state, this paper gives method of judging whether blurring image is uniform defocusing, the method compares picture of real gray with picture of ideal energy, it is clarity and intuitionist. Based on that, author applies a simple algorithm of restoration: counter-filter method, this algorithm is designed according to theory that the Fourier Transfer of two function’s convolution is equal to multiply of two function’s Fourier Transfer. At the same time, brings forward two evaluation-algorithms of judging definition of some restored image according to feature of count-filter, in order to get the best restored image.Uniform linear motion blurring is also a common kind of blurring in motion, because motion of changed speed and non-linear can be regarded as composed result of uniform linear motion. Author applies a high-speed algorithm of restoration: difference method, which complicated course of restoration is converted some simple part and each part is only added to sum. At last, discusses some evaluation-algorithms of judging restored image’s definition, in order to get the best restored image.This paper validates correctness of algorithm and theory through specific restored image, looking into result of experiment, we can know effect is satisfactions,but it still has space to advance.更多还原