ABSTRACT
ABSTRACT: This paper proposes an approach to detect changes between two longitudinal retinal images of the same patient based on multi-attributes fusion. With the approach, two retinal images are firstly aligned spatially by registering them. Then illumination variations between these two images are reduced by employing the modified Single Scale Retinex method. Finally, changes between these two images are detected based on two attributes: intensity and texture. Experimental results show that the proposed approach is effective and robust in identifying small and weak dissimilar variations from retinal images, and has certain advantages compared with some state-of-the-art methods.
