中国科学院光电技术研究所机构知识库
Advanced  
IOE OpenIR  > 光电技术研究所博硕士论文  > 学位论文
题名:
成像系统图像盲复原技术研究
作者: 谢盛华
学位类别: 博士
答辩日期: 2007-05-31
授予单位: 中国科学院光电技术研究所
授予地点: 光电技术研究所
导师: 张启衡
关键词: 成像系统 ; 图像盲复原 ; 湍流退化 ; 运动模糊 ; 离焦模糊
其他题名: The Study on Blind Image Restoration Technologies in Imaging System
学位专业: 信号与信息处理
中文摘要: 图像盲复原是恢复成像系统图像质量、提高图像清晰度的一种关键技术。本文重点针对成像系统中的大气湍流退化图像、噪声干扰下的运动模糊图像以及离焦模糊图像的盲复原技术展开深入研究,针对实时成像系统的应用需求,提出一些新的盲复原方法和技术,并取得了重要进展。 针对大气湍流退化图像的复原,本文以提高图像复原质量、缩短图像复原时间为根本出发点,从多角度对成实际采集的湍流退化图像复原进行了深层次的研究,提出了多种方法和技术途径来解决成像系统湍流退化图像复原的问题: (1) 对一定湍流影响下的短曝光图像,提出了具有G类点扩展函数的假设,从而提出了基于APEX法的湍流退化图像复原算法。该算法利用退化图像的频谱信息来估计G类点扩展函数,并在此基础上恢复清晰图像。本文根据实际采集的短曝光湍流退化图像的频域特征,提出了多方向综合估计G类点扩展函数的新方法,减少了随机因素的干扰,弥补了原有单一方向估计G类点扩展函数不稳定的缺陷,从而增强了APEX法对真实湍流退化图像复原的稳定性及实用性。 (2) 在图像退化参数模型未知的情况下,提出了基于离散点扩展函数估计的湍流退化图像复原算法。该算法利用相邻两帧短曝光图像的频谱信息建立了离散点扩展函数估计的数学模型,并将点扩展函数的估计问题转化成目标函数的极值优化问题。利用点扩展函数的先验约束信息对目标函数进行了改进,大大提高了点扩展函数估计的精度,增强了图像复原的稳定性和可靠性。利用牛顿迭代法对点扩展函数寻优求解,缩短了图像复原的时间。 (3) 根据极大似然估计原理,提出了基于极大似然估计的单帧图像盲复原算法。将原始图像和点扩展函数当作未知参数,建立了极大似然估计复原的数学模型,并将图像复原问题转化成目标函数的极值优化问题。在引入先验约束信息的基础上,提出了正则化技术优化图像复原,明显增强了原有算法的抗噪性和稳定性。利用共轭梯度法进行极值搜索,提出了图像和点扩展函数交叉迭代估计的流程,选择合理的点扩展函数初始估计加快了图像复原的进程。 (4) 根据多通道图像退化模型,提出了短曝光序列多帧图像盲复原算法。利用极大似然估计原理,建立了Poisson模型下的多帧图像复原似然函数,并对其进行了优化改进,利用EM算法极大化似然函数,推出了EM迭代图像复原方法;结合离散点扩展函数(DPSF)估计的原理,提出了一种新算法——基于DPSF估计的多帧图像复原混合算法。新算法利用两帧短曝光图像估计出离散点扩展函数,并以此作为EM法图像复原的初始估计进行迭代图像复原,减少了原有多帧图像复原算法初始估计的盲目性, 增强了原有算法图像复原的收敛性和稳定性,加速了图像复原的进程,同时具有良好的抗噪性能以及超分辨率复原的能力。根据工程应用需求,提出了多帧图像复原算法实现的方案,为算法的硬件实现奠定了基础。 针对运动模糊图像,本文提出了峰值轨迹域运动参数估计法。在受噪声的影响下,将图像频域原点附近受噪声影响相对较小的区域定义为峰值轨迹域,并在峰值轨迹域内计算出峰值轨迹的方向,利用峰值轨迹方向与运动方向的正交关系来估计运动方向,并在此基础上估计运动模糊的距离,从而实现了带噪声的运动模糊图像的盲复原。该方法良好的抗噪性能明显优于传统的方法,具有更强的实用性。 针对离焦模糊图像,本文从图像清晰度评价的角度出发,以圆盘离焦模型为基础,提出了基于峰度最小的离焦模糊图像复原算法。该方法抛开了传统方法直接从频域特征中估计模糊半径的思路,采用一系列模糊半径对噪声干扰下的模糊图像进行维纳滤波复原,并选取峰度最小的复原图像作为最终的复原结果,成功实现噪声干扰下离焦模糊图像的复原。根据调焦原理中的粗调和精调的思想,提出了离焦模糊半径粗搜索和精搜索的新策略,加快了图像复原的进程。 本论文在对成像系统采集的模糊图像分析的基础之上,结合工程实践,提出了多项新的方法和技术。这些技术和方法在实际图像复原实验中,都具有显著的效果,并在一定程度上解决了图像盲复原质量和盲复原处理时间之间的矛盾,为图像盲复原技术真正用于实际成像系统的图像恢复奠定了理论基础。
英文摘要: Blind image restoration is a kind of key technology of restoring image quality and enhancing image definition. According to the requirement of real-time imaging system, the Ph D dissertation deep researched into the blind image restoration technologies of atmosphere turbulence-degraded image, motion blurred image, out-of-focus blurred image in imaging system, presented some new methods and technologies of blind image restoration, and obtained important progress. To atmosphere turbulence-degraded image restoration, regarding enhancement of restored image quality and decrease of restoring time as the essential goal, the paper carried through deep researches from different point of views and brought forward some methods and technologies to resolve the problems of turbulence-degraded image restoration in actual imaging system. (1) It is assumed that some short-exposure turbulence-degraded images have class G point spread function (PSF); accordingly the restoration algorithm for turbulence-degraded images based on APEX method is present. The algorithm estimates class G PSF by the frequency-domain information of turbulence-degraded image, then restores clear image. Considering the frequency-domain features of actually collected turbulence-degraded image, a new method is presented which synthetically estimates class G PSF by multi-direction one-dimension frequency spectrum. The new method decreases the affection of random disturbance, makes up the objection that original single-direction estimation of class PSF is instable. Thus, the stability and practicability of APEX method are greatly enhanced in real turbulence-degraded image restoration. (2) On the condition that the parameter model of image degradation is unknown, the paper put forward the restoration method for turbulence-degraded image based on the estimation of discrete point spread function (DPSF). The method establishes the math model of DPSF estimation by the frequency-domain information of two short-exposure images, and estimates PSF by extremum optimization of an object function. The object function is improved by prior-constraint information, accordingly the precision of DPSF estimation is greatly enhanced, and the stability and reliability of image restoration are built up. It shortens image restoration time that the optimization estimation of DPSF is carried out by Newton iterative method. (3) According to the principle to maximum likelihood, the paper brought forward blind image restoration algorithm of single-frame image based on maximum likelihood. The algorithm regards the original image and PSF as unknown parameter, establishes the optimization estimation model of blind image restoration, and changes the problem of image restoration into extremum optimization of object function. The image restoration is optimized by regularization technologies and priori-constraint information, which obviously enhances the stability and the capability of resisting-noise. Searching extremum by conjugate grads method, the algorithm present cross iterative estimation flow of image and PSF. An appropriate initial estimation of PSF accelerates the course of image restoration. (4) According to the multi-channel image degradation model, the paper studied multi-frame images blind restoration algorithm for short-exposure sequence images. The maximum-likelihood function of multi-frame images restoration is established according to maximum-likelihood theory and Poisson probability model, and is improved by prior information. The maximum-likelihood function is maximized by EM algorithm; accordingly, EM iterative image restoration method is deduced. On basis of EM image restoration, combining the principle of estimating DPSF, the paper present a kind of blend method of multi-frame image restoration based on DPSF estimation, which estimates DPSF by two sequential images, and regards the DPSF as the initial estimation of EM image restoration, accelerates the course of EM image restoration, and have strong resist-noise capability and super-resolution performance. According to the requirement of engineering application, the realization project of multi-frame image restoration algorithm was designed, which established a firm foundation of hardware realization. Concerning the motion blurring image with noise disturbance, the paper presented an algorithm of estimating motion parameter in the peak-trace domain. On condition of noise disturbance, the neighborhood of frequency-domain origin is defined as peak-trace domain, and the peak-trance direction is extracted in the peak-trace domain. According to the perpendicular relation between motion direction and peak-trace direction, the motion direction can be obtained. On the basis of motion direction, motion blurring length can be estimated. Consequently, motion blurring image with noise disturbance can be restored. The algorithm has better resist-noise performance and stronger practicability than traditional methods. As for out-of-focus blurred image restoration, on the basis of disk out-of-focus model, the paper presented the blind image restoration algorithm based on kurtosis minimization. The algorithm gives up the idea of traditional method which estimates blurring radius by frequency-domain features. It restores blurred image by Wiener filter according to some blurring radius, and regards the restored image with minimum kurtosis as the final restored image, and successfully realizes the image restoration of out-of-focus blurred image with noise disturbance. According to the principle of rough focusing and accurate focusing, the paper put forward a new strategy of rough searching and accurate searching of blurring radius. Consequently the course of image restoration is greatly quickened. According to the features of image in actual imaging system and requirement of actual engineering application, the paper brought forward some new methods and technologies of blind image restoration, which are proved to be effective in actual image restoration experiment, resolved the contradiction between blind restoration quality and restoration time to some extent, and established the theory basis by which blind image restoration technologies were applied into actual imaging system.
语种: 中文
内容类型: 学位论文
URI标识: http://ir.ioe.ac.cn/handle/181551/222
Appears in Collections:光电技术研究所博硕士论文_学位论文

Files in This Item:
File Name/ File Size Content Type Version Access License
10001_200418015127016谢盛华_paper.doc(9604KB)----限制开放View 联系获取全文

Recommended Citation:
谢盛华. 成像系统图像盲复原技术研究[D]. 光电技术研究所. 中国科学院光电技术研究所. 2007.
Service
Recommend this item
Sava as my favorate item
Show this item's statistics
Export Endnote File
Google Scholar
Similar articles in Google Scholar
[谢盛华]'s Articles
CSDL cross search
Similar articles in CSDL Cross Search
[谢盛华]‘s Articles
Related Copyright Policies
Null
Social Bookmarking
Add to CiteULike Add to Connotea Add to Del.icio.us Add to Digg Add to Reddit
文件名: 10001_200418015127016谢盛华_paper.doc
格式: Microsoft Word
此文件暂不支持浏览
所有评论 (0)
暂无评论
 
评注功能仅针对注册用户开放,请您登录
您对该条目有什么异议,请填写以下表单,管理员会尽快联系您。
内 容:
Email:  *
单位:
验证码:   刷新
您在IR的使用过程中有什么好的想法或者建议可以反馈给我们。
标 题:
 *
内 容:
Email:  *
验证码:   刷新

Items in IR are protected by copyright, with all rights reserved, unless otherwise indicated.

 

 

Valid XHTML 1.0!
Copyright © 2007-2016  中国科学院光电技术研究所 - Feedback
Powered by CSpace