Our preliminary studies have demonstrated that pre-processing with edge-detection algorithms allow for registration of photographic and angiographic data where the vessel gray levels may vary from black to white (Figure 2). Early non-real-time studies demonstrating the applicability of edge-detection-based algorithms for image registration were performed on a PC compatible computer with Matrox Inspector (Quebec, Canada) software.
For potential near-real-time application, the biomicroscopic image will be acquired and edge detected. A pruned search tree will then be searched extending from no displacement to the maximum displacement as determined by the frame rate and image size. The Hausdorff distance will be minimized, with computational frugality achieved by a) the small size of the illuminated biomicroscopic image, b) calculations with binarized images (permitting logical operations), c) restricting our template search to translational displacements with correction for scaling, and d) limiting the search area. Crucial to efficient application is the off-line montage creation, allowing for instant access to photographic and angiographic data. Optimum temporal performance may be achieved with a small illuminated fundus area, i.e. a small template, with the remainder of the image presented efficiently from previously stored data. Along these lines, the influence of the size of the illuminated biomicroscopic image on the speed and fidelity of registration will be explored.