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:
基于自适应光学获得的眼底图像拼接算法研究
作者: 刘玉然
学位类别: 博士
答辩日期: 2009-05-26
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: 张雨东
关键词: 人眼眼底图像 ; 自适应光学 ; 图像拼接 ; RANSAC ; 马氏距离 ; Harris ; 角点匹配 ; KLT ; 多幅图像拼接
其他题名: The Research of the Mosaic Arithmetic of Human Retina Images by Adaptive Optics
学位专业: 信号与信息处理
中文摘要: 眼底循环障碍疾病以及全身性疾病可以不同程度的显征于视网膜。因此,眼底检查不仅对眼科疾病,而且对其他系统及全身性疾病的诊断也有着重要的意义。由于人眼像差的存在,使一般的相机很难为眼科医生提供全面准确的眼底信息,并辅助医生诊断和定位病灶。自适应光学解决了这个问题,它通过校正像差,能够拍摄高分辨率的眼底微细毛细血管。但是由于视野狭小,单幅照片只能反映很小区域的眼底信息,而手动拼接效率低且拼接结果存在明显的拼缝,这样就给医生的诊断造成了困难,因此有必要对自适应光学眼底图像自动拼接算法进行研究。 毛细血管在视网膜上呈立体网状分布。由于眼底呈球面和人眼的抖动影响使得拍摄到的图像不在准确同一层上,这样获得的图片存在着图像重叠区域细节不一致的问题,同时焦距的改变使得图片也存在着尺度变化,自适应光学技术自身的成像特点使获得的图像中间亮周围暗,以及视网膜各区域结构不同造成的反射率不同使得获得的图像灰度不一致,眼底的球面特征也使图像间存在着微小的旋转,这些特点使得自适应光学技术获得的高分辨率人眼视网膜毛细血管图像的拼接十分困难;除此之外,169幅多幅多行图像拼接更增加了拼接的难度。因此,本课题所面临的任务是对多幅多行的,具有尺度变化,旋转,灰度不均匀,细节不一致等特点的自适应光学获得的眼底图像进行拼接。 课题主要分为两部分:第一部分是针对两幅图像拼接的探索;第二部分是对于多幅图像拼接出现变形严重问题的时候,对问题深入分析,并给出解决办法,最后实现。具体如下:首先提出一种基于特征点的图像拼接算法,该方法以改进Harris角点检测算法检测出特征点,以RANSAC角点粗匹配,并选择马氏距离对角点匹配对精确提纯,然后用得到的精确角点匹配对进行仿射参数计算,进而得到两副拼图像接后的结果。但当用这种方法进行多幅图像连续拼接的时候,产生变形严重的问题,无法实现所有眼底图像的自动拼接。经过分析这是由于角点匹配和变换参数计算的不准确引起的。针对角点匹配和变换参数计算的不准确问题,采用了KLT角点检测方法,用傅立叶变换的波形相似性来进行角点提纯,并LM算法迭代计算投影变换矩阵加以解决。最后用逆向变换解决拼接黑洞的问题,用过渡函数对图像间的拼缝进行平滑处理,完成图像融合,获得了较好的拼接结果。本文主要研究内容和成果包括以下几个方面: (1)提出适合人眼眼底图像的角点检测算法——改进的Harris检测的方法,作为人眼眼底图像的角点检测的方法。Harris角点检测方法是一种现阶段应用广泛,角点检测精度高,运算量相对较小的优秀角点检测算法,但是它对于旋转和加噪图像的角点响应值不稳定限制了它的应用,改进的Harris检测方法针对这些缺陷进行改进,使它具有更好的检测性能,更适合于人眼眼底图像的角点检测。 (2)提出适合两幅人眼眼底图像的特征匹配算法。针对人眼眼底图像细节对应不一致、灰度不一致、旋转、缩放使得特征点无法一一对应的问题,提出先用RANSAC角点粗匹配算法再用马氏距离对人眼眼底图像的精确角点匹配对的提纯算法,使得在较小计算量的情况下获得较为精确的特征匹配,大大提高了角点匹配的速度和精度。 (3)对多幅图像的拼接提出了好的解决办法。算法采用了KLT角点检测方法,用傅立叶变换的波形相似性来进行角点提纯,并LM算法迭代计算投影变换矩阵,实现了人眼眼底图像的拼接。 (4)解决了多幅图像的融合问题。针对不规则图像的融合问题首先将其简化成两圆重叠区域的融合问题,并提出两两圆域拼接的融合算法,并获得较好的效果。 大视野,高分辨率人眼眼底图像拼接的完成,为疾病的早期诊断提供了真实可靠的依据,不但可以预防很多致盲疾病的形成,而且对于很多全身疾病和神经系统的疾病预防都有着重大意义。 从图像拼接角度来讲,不但解决有旋转、变形、灰度不一致、细节不对应的多幅多行图像的拼接问题,而且还解决了多幅多行图像之间的融合问题。这种算法可以推广开来,适用于存在微小尺度变化、旋转和不确定平移量,图像本身对应性不是很强的多幅多行图像的拼接和配准中。
英文摘要: Fundus abnormalities caused by ocular and systemic diseases can be shown in retina. Therefore, the examination of the fundus is an applicable tool in diagnosing and monitoring both eye diseases and whole body conditions. Due to the existence of aberration, retina images which are acquired by the common camera can not be utilized to help ophthalmologist on focal location and diagnosis. Adaptive Optics (AO) technique solved this problem by revealing the smallest capillary vessel of the retina through compensating the aberration. However, the field of view is too small to provide enough information in the very small retina region revealed by one AO retina image. Moreover, it is inefficient and ineffective to mosaic artificially, and there are obvious marks in it . Hence, it is essential to carry out research on automatic mosaic technology for the AO retina images. The distributing of capillary vessel in retina is solid reticulation, hence, the images are not on the exactly same layer because of spherical surface of the funds and the aberrations of the eyes, which made the inconsistent details in the overlap area; meanwhile, the scale variety of the images is also existed because of the change of the focus; the characters of AO made the images bright in the middle area and dark in around area, reflectivity of retina is difference in different area, which made the images grey discordance; the images also have toothful rotation because the spherical surface of the funds ; so it is difficult to make the stitching of the high resolution human retina capillary vessel images by AO. Besides, 169 mosaicing multiple and multi-row images increase the difficulties severely. Then, the task of the project is the mosaic of the 169 AO retina multiple and multi-row images, which have scale variation, rotation, grey discordance and detail not corresponding. There are two parts in this thesis: one part is searching mosaic method of two images, the other part is analysing the distortion of multi-images mosaic, and a method is proposed. A robust stitching arithmetic based on feature points is presented. This method is based on the corner points which is detected by improved Harris arithmetic, then, initial point matching by RANSAC, accurate point matching by Mahalanobis distance, at last, affine alternate parameter are calculated, then the two images are stitched. However, the method mentioned aboved is produced seriouse problems when multi-images stitching is carrided through. All the images’ automatic stitching is not completed, the reasons are the imprecision of corner corresponding and transform parameter calculation. To the problem of imprecision of corner corresponding and transform parameter calculation, KLT arithmetic is adopted, after that, Fourierism’ similition as accurate point matching arithmetic, projective alternate parameters are calculated by LM iterative arithmetic. At last, the backward distortion and cap function are adopted to fuse the images,then the good results are obtained. The main task and achievement mentioned above are summarized as follows: (1) A corner detection arithmetic is proposed. Harris’ corner detection method is a widely used excellent detection method characteristed with high precision and low calculationn cost, but it is limited by the unstability for images after rotation and being added by noise. We propose a new corner detection method which is propitious to detection the accurate the corner of the OA retina image. (2) A corner points matching arithmetic is presented. For the characteristics of retina images, such as scale variety, noising, grey discordance, detail not corresponding, etc, a corner points matching arithmetic is presented, initial point matching by RANSAC, accurate point matching by Mahalanobis Distance distance, which can get accurate corner point matching by a few calculation. (3) A method is presented for multi-images stitching. It is based on KLT corner detection, using Fourierism’ similition as accurate point matching arithmetic, and projective alternate parameter are calculated by LM iterative arithmetic, and this method is implemented. (4) The problem of the confusing of multi-images is solved. Firstly, the problem is predigested to the confusing of round images overlaped with each other, a settled method is proposed, and good result is achieved. The well mosaicing of large vision and high resolution human fundus images provide true and reliable evidence for early diagnosis of disease. This technique can not only prevent many blinding disease, but also have significance for precausion of many systemic and nervous diseases. From the point of image mosaicing technique, it is also new to solve the problem of mosaicing multiple and multi-row images with rotation, transformation, inconsistent intensity and details. This not only solves the problem of accumulative error of mosaicing multiple and multi-row images, but also solves the problem of fusing multiple and multi-row images. This algorithm is applicable to the mosaicing and registration of multiple and multi-row images which have small scale variance, rotation, unpredictable translation and lower correspondence between images.
语种: 中文
内容类型: 学位论文
URI标识: http://ir.ioe.ac.cn/handle/181551/298
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_200518015127037刘玉然_paper.pdf(7632KB)----限制开放View 联系获取全文

Recommended Citation:
刘玉然. 基于自适应光学获得的眼底图像拼接算法研究[D]. 光电技术研究所. 中国科学院光电技术研究所. 2009.
Service
Recommend this item
Sava as my favorate item
Show this item's statistics
Export Endnote File
Google Scholar
Similar articles in Google Scholar
[刘玉然]'s Articles
CSDL cross search
Similar articles in CSDL Cross Search
[刘玉然]‘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_200518015127037刘玉然_paper.pdf
格式: Adobe PDF
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