【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 ,且具有运算速度快的特点更多还原

【Abstract】 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.更多还原