Retinal Vessel Extraction Method and System Based on Dynamic Scale Allocation

Abstract

The invention discloses a method and system for extracting retinal blood vessels based on dynamic scale distribution; steps of the method: retinal image preprocessing: performing contrast enhancement on the green channel component of the color retinal image; image segmentation: dividing the preprocessed retinal image into Set the number of sub-images; blood vessel classification: divide the blood vessels in each sub-image into three categories: large, medium and small; dynamic scale allocation: dynamically select filters of different scales to enhance blood vessels of different widths; multi-scale matching filtering; threshold processing: The vascular structure is extracted and the non-vascular structure is eliminated, and the extraction results of all sub-images are reassembled to obtain a binary image of the retinal vascular network; a retinal vascular network image with high segmentation accuracy is obtained after post-processing. The beneficial effect of the present invention is to realize the extraction of blood vessels from retinal images, avoid overestimation of the width of blood vessels while eliminating complex non-vascular structures, and realize simpler and more accurate extraction of retinal blood vessels.

Description

technical field [0001] The invention relates to a method and system for extracting retinal blood vessels based on dynamic scale allocation. Background technique [0002] So far, the commonly used retinal vessel automatic extraction algorithms are: [0003] 1. Algorithm based on retinal blood vessel tracking method. This type of method can extract the retinal blood vessel network relatively completely, but the complexity of the algorithm is high and the amount of calculation is large. In addition, for some retinal blood vessel images with low contrast, the extraction accuracy of such algorithms is not enough. Among them, the typical retinal blood vessel tracking algorithm is based on the fuzzy C-means clustering algorithm proposed by Tolias in 1998, which selects a suitable seed point at the beginning of the blood vessel (optic disc), and thus performs the whole retinal blood vessel network. track. Establish a one-dimensional model of the retinal blood vessel cross-sectio...

AI Technical Summary

Problems solved by technology

However, this method still has certain limitations: it uses three filters of different scales to extract retinal blood vessels. For the blood vessel widths that vary greatly, there are still some blood vessel widths that cannot be well enhanced. Therefore, It is difficult to accurately segment all blood vessels; in addition, for some retinal images with lesion areas, there will be wrongly segmented pixels

Images