一种新型的视盘分割方法

A Novel Optic Disc Image Segmentation Method

视盘的检测对于眼底图像分析和计算机辅助诊断视网膜神经疾病有重要意义。论文提出了一种新型的视盘分割方法,新方法的思想是在去除血管的背景图像应用梯度向量场GVF活动轮廓模型。首先,从原始图像获取部分背景图像的像素,即获得在去除图像中血管像素之后的部分像素,然后利用取得部分背景的像素重塑整个背景图像,最后在恢复的背景图像中应用GVF活动轮廓模型分割出视盘,GVF Snake模型的初始曲线为Hough circle。选取了不同的眼底医学图像进行了大量的实验,结果表明论文算法的分割结果十分准确,具有较好的分割效果。

The optic disc image segmentation is very important for fundus image analysis and computer-aided diagnosis of retinal optic nerve disease. In this paper, a novel method for optic disc localization is presented by applying gratitude vec- tor field（GVF） snake model to a background image without vessels. Firstly, the vessels are filtered from the original image and the partly background is obtained. And then a fitting method is used to recover the whole background image without ves- sels. This background image gets rid of the influence of complex blood vessels and other barriers and makes the optic disc easier to be detected. Finally GVF snake model is applied to detect the optic disc from the background image with an initial Hough circle. Plenty of experiments on different medical fundus images show that the new-presented algorithm has a better segment results than the methods using the GVF snake model directly.