The Vessel Tree Segmentation for Retinal Image via Matched Filter Combining Enhanced Frame

Heng Dong, Lifang Wei

Abstract: The vessel tree segmentation for retinal image is most conducive to check fundus diseases. But the automatic vessel tree segmentation is still a challenging issue in the presence with computer assist. To tackle it, a vessel tree segmentation frame is proposed, which combines the matched filtering and image enhance processing to complete the retinal vessel tree segmentation for color retinal images. According to the difference of three-component RGB of color image, the green channel is utilized to obtain graying image and enhance the contrast with CLAHE. The morphological Top-hat and Bottom-hat is used for further retinal image preprocessing. Then the matched filter is used to segment the vessel tree that is based on Gaussian kernels. Finally, we have extensively compared with multiple variants of matched filter frame method through experiments, which show that our frame is efficient.