Retinal Image Automatic Registration Based On Local Bifurcation Structure

Retinal images can be used for helping ophthalmologist to diagnose the eye diseases, and glaucoma, hypertension, coronary heart diseases and so on, so it's very important. The most important content of this article is a new structural feature for retinal registration. There are many methods of retinal image registration; some are also used by this time. Lots of methods used the main bifurcation points and angles. This method is well and effective, but there are some problems, for example, the angles about bifurcation maybe repeat, so in two images, the bifurcation registration may be failed. On the basis of this, in this paper we proposed a structure-matching registration method. The bifurcation structures are consists of a main bifurcation point and its three connected branches. The each numerical value of bifurcation is composed of the bifurcation angles and its three lengths by normalized. On the other hand the bifurcation structure will be invariant against rotation, translation and scaling. By this method, we can reduce the matching error caused by repeated. This method could be used alone, also could be used with other methods.