基于ICP和SVD的视网膜图像特征点配准算法  

New Method for Automatic Retinal Images Registration Combined with ICP and SVD

作　　者：张二虎 [1] 卞正中 [2] 张燕 [3] 段敬红 [3] 吴学毅 [3] 

机构地区：[1]西安交通大学,生命科学与技术学院,陕西,西安,710049,西安理工大学,信息科学系,陕西,西安,710048 [2]西安交通大学,生命科学与技术学院,陕西,西安,710049 [3]西安理工大学,信息科学系,陕西,西安,710048

出　　处：《小型微型计算机系统》2004年第10期1811-1813,共3页Journal of Chinese Computer Systems

基　　金：国家自然科学基金 ( 60 0 710 2 9)资助

摘　　要：视网膜图像配准是视网膜眼底疾病诊断及激光治疗中的关键一步 .针对荧光造影视网膜图像和无赤光视网膜图像的特点 ,提出一种采用迭代最近点 (ICP)和奇异值分解 (SVD)的方法 ,用于视网膜图像之间的配准 .即用 ICP算法确定两个特征点集的对应性及点集数目不等的问题 ,用 SVD方法求解空间变换参数 .实验证明在一个点集数目缺少75 %的情况下 ,算法仍然能达到较好的配准精度 ,可以满足荧光造影和无赤光视网膜图像之间的配准 。Retinal images registration is an important step in retinal fundus diseases diagnosis and laser treatment in order to combine different retinal images information, such as fluoroscein angiography image (FA ) and red free image(RF). A new method combined with ICP and SVD is proposed for retinal images registration. Retinal vascular branch points and cross points are used as feature point set and the feature points correspondence in RF and FA image as well as the difference in amount are solved by ICP method, while the transform parameters is solved by SVD method. The experiment result shows the proposed method can register points sets precisely even on loss of 75 percent points in one point set and performs very fast.

关 键 词：图像配准 视网膜图像 特征点对应 奇异值分解 

分 类 号：R318.04[医药卫生—生物医学工程]