A Hybrid Morphological Based Segmentation Method for Extracting Retina Blood Vessels Grid

Abstract

The patterns of retinal blood vessels play major role in many different applications, such as diseases diagnosis and human identification. The accurate segmentation of vessels body appeared in retina images is vital to make successful human identification decisions. This paper presents a new method for vascular network extraction from color retinal images. The proposed method consists of three main stages: Preprocessing, segmentation, and post-processing. Preprocessing stage is applied to enhance the local appearance of blood vessels in retinal images; its main task is to make compensation for the global/local contrast variance over all parts of the retina area, such that the dynamic range for brightness levels of the vessels' pixels becomes narrow and lies in the dark region of brightness scale. In segmentation stage, the grid of retina vessels had been extracted using thresholding method; where the vessels appear dark, thin and connected bodies in retina area. Finally, post preprocessing stage is applied to eliminate the noise and to remove the produced disconnections in the extracted vessels due to thresholding. The proposed method was tested on the two publicly available datasets: (i) DRIVE (Digital Retinal Images for Vessel Extraction) and (ii) STARE (Structured Analysis of the Retina). The test results indicated that the proposed method is efficient to segment the large vascular areas and outperforms of many introduced methods in the literature. The test results indicated that the attained accuracy of the proposed method was 97.41% in DRIVE dataset, and 97.43% in STARE dataset