Segmentation of a class of ophthalmological images using a directional variance operator and co-occurrence arrays

The posterior capsule opacification images considered are images of the membrane encapsulating an artificial lens implanted during cataract surgery in place of the natural lens. The images are taken to monitor the state of the patient’s vision after the surgery. Subsequent to the surgery, the membrane of the posterior capsule may become opacified, thus degrading the patient’s vision. We discuss the methodology used and the results obtained in the segmentation of the images into transparent and opacified regions. The opacification is primarily characterized by its texture, therefore a directional standard deviation operator is applied to an image giving rise to a family of ‘‘conjugate’’ images. From these images, the multi-dimensional histogram (co-occurrence) array is calculated and subsequently approximated by Gaussian distributions to form the basis for the segmentation step.