基于知识的眼底荧光血管造影图像序列配准新方法

A NOVEL AND KNOWLEDGE-BASED METHOD FOR AUTOMATED REGISTRATION OF FLUORESCEIN RETINAL ANGIOGRAMS

本文提出一种用于荧光眼底血管造影图像序列快速自动配准的新方法。首先对序列图像进行采集及前处理,然后进行基于知识的特征提取及特征筛选,再对图像配准参数进行求解及优化。分别采用本文方法及传统方法进行了对比实验与结果比较。较之经典方法,本文方法具有快速、配准精度高及抗干扰能力强的优点。

Comparisons of flurescein angiograms at different phases are essential in monitoring the progress of many systemic diseases and in assessing the response to treatment. Accurate image registration is a prerequisite for quantitative estimates of changes at specific locations. In this paper, a novel and computationally efficient approach to an automated registration of fluorescein retinal angiograms was put forward. First, sequential images were captured and preprocessed (WT-based de-noising). Then, knowledge-based vascular features detection and sifting methods were applied. Finally, image transformation was developed and refined. The performance of the method was compared to that of the conventional methods and found it to be superior. Experment results show that retinal images can be fast registrated by the proposed method with high accuracy on conventional computers and with an outstanding robustness against correlated noise and disturbances, such as those encountered with non-uniform, time varying fluorescent' intensity.