A feature-based image registration method is presented for the use on the multispectral skin images. The proposed method allows the use and combination of different kinds of feature points, so it can be adjusted to different applications. The local minima and maxima are used as features for the skin images. By analyzing all possible shifts among feature points in the reference and sensed images, the proposed method efficiently finds translation among these images. The method is expanded to also determin e rotation and scale. The proposed method is compared with other area-based and feature-based methods (proposed in the literature specifically for skin images) by testing those methods on a set of multispectral skin images.

Kadikis, R. (2015, April). Registration method for multispectral skin images. In 2015 25th International Conference Radioelektronika (RADIOELEKTRONIKA) (pp. 232-235). IEEE.

pdf