Fundus Image Mosaic Generation For Large Field of View

Abstract

Considering the importance of retina in the eye as many ocular and systemic diseases are manifested in the retina, the need for detailed study of 2-D retinal images, 3-D retinal images and panoramic retinal images is increasing rapidly. Many structures in human body are responsible for the process of vision but our works is strictly related to the retina as most diseases of eye manifest themselves in retina of the human eye. Digital fundus images are commonly used for computer aided diagnosis of different eye diseases such as diabetic retinopathy, glaucoma and age related macular degeneration. A very important problem with fundus cameras is that they provide fundus image only for a small field of view (FOV). Our thesis work presents a novel method to increase the FOV by stitching different fundus images of a particular patient. The proposed system uses weber local descriptor (WLD) based descriptors and generates a blended image by combining all available images. This work also compares the proposed system with Harris corner detector, SURF, SIFT and ASIFT based descriptors for the same application. A local dataset of 15 patients with different number of fundus images is gathered for proper validation of the proposed system