Abstract: An edge detection algorithm which is applied to anterior chamber OCT（Optical Coherence Tomography） images has been proposed. The algorithm uses multi-structure morphology elements to detect edges first, and then fuses these obtained multi-structure edge images by dynamic adaptive weight according to the edge pixel’s gray-scale value distinctiveness. The finally edge image is obtained after erasing some smaller interference areas, which are detected by counting the areas of connected domain in the fusion image. The pre-knowledge of positions of the corner points have been used to detect the feature corner points of the cornea. The simulated results have shown that the proposed algorithm can effectively avoid the occurrence of mutational pixels in the OCT image edge results, erased interference area; so more clear edges and high accuracy corner points of the cornea can be obtained compared to traditional edge detection algorithms.

Key words: Optical Coherence Tomography（OCT） image, anterior chamber, adaptive weighted fusion, multi-structure morphology elements, edge detection, connected domain, feature corner points

摘要： 提出了一种适用于眼前节组织OCT图像的边缘检测算法。该算法在单尺度下用多个结构元素进行边缘检测，根据边缘图像灰阶值的差异性，采用动态自适应权重进行像素点融合；再利用连通域的方法抹去面积小的干扰区域，最终得到多结构元素单尺度边缘检测图像，并在其上通过象限区间有效地提取出了角膜特征角点。仿真结果表明边缘特征明显，较以往边缘检测算法有效避免了OCT图像边缘结果的突变像素点的出现，抹去了干扰区域。因此，提出的特征角点具有较高的准确性。

关键词: OCT图像, 眼前节组织, 自适应权重, 多结构, 边缘检测, 连通域, 特征角点