Medida de similitud combinada para el registro y fusión de imágenes

Abstract

The aim of this paper is to present a combined similarity measure for non-rigid image registration and fusion based on normalized cross correlation and mean square difference. Many similarity measures are used to image registration, but there is not a single measure that can produce best results on all images. The use of combined similarity measures allows conjugating the detection properties of involved measures. The combined similarity is used to multimodal medical image registration with a voxel based property method. For that purpose, firstly, the algorithm compute the pixel intensity relation between images to register with a non-lineal estimation method based on radial basis functions using joint grey value histogram. Then, non-rigid motion estimation method based on fuzzy models is used to image registration. The transparency technique is used to fuse the images. Multimodal registration of a computed tomography image and a positron emission tomography image with non-rigid motion are presented.