A novel retinal vessel extraction method based on dynamic scales allocation

Abstract

Automatic extraction of retinal vessels is significant for diagnosis of eye diseases. Currently, the automatic extraction of the vessels in the retinal images with very low contrast and various widths is a bottleneck. In this paper an effective retinal blood vessel extraction method to detect fine vessels more accurately was presented. The contribution of this work is that a novel dynamic scale allocation scheme of the matched filter was proposed. The whole image is divided into sub-blocks. The histogram of each sub-block is fitted by Gaussian function whose fitting parameters are used to select the scales. Compared with the existing blood vessel extraction using uniform multiscale matched filter, the proposed method detects many fine vessels drowned by noise and has good width estimation.