| Resumen |
The fundamental aim of this work is to achieve the multimodal image registration of two retinal images of different types. The performance of the proposed methodology was tested on a set of pairs of retinal images, with and without microaneurysms. Each pair consists of a color optical and a gray-level fluoresceinic image of the same retina, which have very different radiometric characteristics. In the process it is used the scale-invariant feature transform methodology (SIFT). The initial and main stage is the preprocessing stage that allows the next four stages, namely, features detection; feature matching, selection of characteristics and transformation. With the uncontrolled quality of images, it was possible to register 26 pairs out of 35 pairs of retina images, which corresponds to 74.28% of the total. From the registered 26 pairs, 14 pairs, which correspond to 53.85%, were retinal images without microaneurysms, and 46.15% were retinal images with microaneurysms. |
