Detección y descripción de puntos característicos en imágenes multiespectrales utilizando esquemas clásicos

Abstract

In this work we study the behavior of algorithms used in computer vision for key point detection and description when they are used in multispectral imaging. In this study we search for those variables of SIFT and SURF algorithms that can be adapted in order to have more key points as well as to increase the number of correspondences between images of the visible spectrum (VS) and images of infrared spectrum. Through this document we will present the variables that were adapted, improving the key point detection stage by increasing the number of points detected in the infrared spectrum; unfortunately, this improvement was not in the description stage, since key point correspondence between VS-IR remain lower.