Current improvements in interpretation of posterior capsular opacification images

Abstract

We present an application of curvature-driven min/max flow and anisotropic diffusion in processing posterior capsular (PCO) images. PCO images present the back surface of the lens implanted during cataract surgery and are used to monitor the state of the patient's vision. Our standard segmentation technique which is based on the variance based co-occurrence matrices often requires an enhancement of variance images prior to segmentation. A number of enhancement methods are based on partial differential equations, and we present two such methods. We demonstrate that the curvature-driven flow seems to enhance better the significant edges in the image, whereas the anisotropic diffusion seems to work better with smoothing intra-regional image features.