基于稀疏表示的视网膜眼底图像血管检测

Detection of Blood Vessels in Retinal Fundus Images Based on Sparse Representation

一般情况下,通过在血管中注入造影剂在X光下所显示的影像来检测血管会受背景噪声、血管形态模糊和目标混杂等影响,鉴于此,提出一种基于过完备字典的图像稀疏表示的血管检测方法。该方法先建立一个关于血管图像的背景图像的过完备字典,然后用SRC算法会得到一个系数矩阵,并利用系数矩阵与字典矩阵相乘,便能重构出一个底板图像。为了突出血管,只需将重构的底板图像与原图像做相减,这样得到的新图像就能去除噪声,使血管能在视网膜眼底图像中更清晰地显示出来。实验结果表明,该方法对于视网膜眼底图像处理效果良好,使视网膜眼底图像中的血管能更容易被检测到。

Generally, a developer was injected into the veinsdisplayed under X light imagingto detect blood vessels would be affected by the background noise,vascular morphology and target fuzzy hybrid. Therefore, a new method of image sparse representation based on over complete dictionary is proposed to detect blood vessels.Firstly,the method establishes a over complete dictionary of the background image of the blood vessel.Next,we will get a coefficient matrix by SRC algorithm, then multiplying the coefficient matrix with the dictionary matrix, the image of a bottom plate can be reconstructed. In or- der to highlight the vessel, as long as the reconstruction of the bottom image and the original image subtraction, this new image will be able to remove the noise, it would make blood vessel in retinal fundus images more clear displayed. The ex- perimental results shows that this method has a good effect for the treatment of retinal fundus image, so that the blood vessels can be more easily detected in the retinal fundus images.