基于相位相关和非刚性光流估计的眼底ICGA图像精细拼合方法

ICGA Fundus Image Mosaicing Based on Phase-correlation Method and Nonrigid Optical Flow Estimation

眼底循环障碍疾病以及全身性疾病导致的眼底改变均可不同程度地显征于眼底图像。在 ICG眼底造影结合共聚焦激光显微镜技术条件下 ,可得到高分辨率的眼底视网膜和脉络膜局部图像 ,而眼底病症可能散布于眼底各生理部位 ,眼底图像序列的自动拼接可解决分辨率和单次可观测视野间的矛盾。采用相位相关法匹配 ICGA眼底图像 ,应用基于遗传算法的非刚性光流场估计方法计算未建模的眼底微小运动 ,配准不完全匹配局部。文中方法完备了 ICGA眼底图像拼接方法 ,提高了精度和正确匹配率 ,实验中有效地克服图像虚影 ,保留了图像细节 。

Fundus imaging has been used to monitor retinal and choroid manifestations of various diseases both in eye and other parts of human body frequently.Indocyanine Green angiography under scanning laser ophthalmoscopy is applied to obtain high quality fundus images; furthermore,image mosaic is built to extend field of view without lowering the image resolution.Phase correlation method is adopted to find the image translation coarsely first,then the overlapping parts is cropped,and the optical flow between them is computed based on genetic algorithm,according to which the image pairs is warped locally to achieve better registration,thus the fine mosaic of fundus is obtained.The method is proved to be robust by experimental results.