An Approximate Skeleton Frog Group Numbering Method for Clustering and Segmentation of Fundus Vascular Images

Abstract

The invention relates to the technical field of fundus blood vessel image clustering operations, in particular to an approximate skeleton frog group numbering method for clustering and segmentation of fundus blood vessel images. The present invention uses a clustering method to segment and process the fundus image, and locates and eliminates the lesion points according to the characteristics of the lesion points being highlighted. In order to obtain a better clustering segmentation effect, the K-means algorithm is improved by using the hybrid leapfrog algorithm, which is more effective and easy to understand in the intelligent algorithm, and the approximate skeleton is used to further make full use of the local optimal solution obtained by the algorithm. The improved algorithm It can effectively overcome the shortcomings of the original K-means algorithm that is easy to converge to the local optimum and cannot effectively perform image segmentation, obtain better fundus blood vessel clustering and segmentation results, and more accurately separate the fundus blood vessel lesion points.

Description

technical field [0001] The invention relates to the technical field of fundus blood vessel image clustering operations, in particular to an approximate skeleton frog group numbering method for clustering and segmentation of fundus blood vessel images. Background technique [0002] The retinal fundus image is mainly composed of the fovea, macula, blood vessels and optic disc. This image is of great significance for the diagnosis of diseases including diabetes, glaucoma, coronary heart disease, etc. Therefore, the processing of the fundus image is very important. In order to reduce the workload of ophthalmologists, facilitate physicians' judgment, and provide methods for large-scale image processing, the application of machine vision in medical images has attracted the attention of many researchers. [0003] In fundus images, blood vessels are an important part, and their structural and morphological changes can effectively point out the development of many systemic diseases i...

AI Technical Summary

Problems solved by technology

However, many blood vessel segmentation algorithms, including matched filtering, rely more on the color difference between pixel blocks, so it is easy to treat some large highlighted lesion blocks caused by diseases as blood vessels, which reduces the accuracy of blood vessel segmentation. , and waste a lot of resources

Images