Abstract

Retinal images play a major role in several applications including the ocular fundus operations and human recognition. They also play vital roles in detection of diabetics in early stages by comparing the states of retinal blood vessels. The detection of blood vessels from the retinal images is usually a tedious process. Most of the research works confirms that, detection result of retinal blood vessels serves some diseases by predicted at an early stage. In our previous research work, Retinal Image Contrast Enhancement Using Nonsubsampled Contourlet Transform (NSCT) extracts the geometric information of images, which can be employed differentiate noises from weak edges. However, in this method the weak edge enhancement have some losses. These issues are solved in our proposed research method by introducing Dynamic histogram equalization (DHE) technique which performs the enhancement of weak edge vessels of a retinal eye image makes without any loss. This method used to detect the blood vessels efficiently and gives better quality of the retinal images with textures, clarity and free from noise. After enhancement of weak edge vessels of retinal eye, then the same procedure followed in our previous work using morphological operations is carried out to remove false edges. Then introducing Hybrid fuzzy morphological and connected components labelling (HFMCCL) method is used to detect and counting the number of vessels and length of the vessels in retinal images are identified in less computational time. Finally a small vessel tracking and reconstruction process is used for tracking the centreline of the blood vessels, after extracted vessels are reconstructed with pixel painting. Hence, the proposed approach gives high performance in terms of reliable accuracy in detecting the blood vessels than the previous approach. Experimentation is conducted in the Matlab simulation environment and it is confirmed from the simulation outcome.

Paper Details