Descritor de imagem baseado em curvas de Hilbert

Abstract

The feature description is an essential part of image processing to achieve good results in feature detection techniques. Besides, it is one of the reasons for the progress in visual recognition in the last decade. Although many papers in computer vision have been published since then, there is a lot of opportunity for improvement of current models and new contributions to this field. This paper proposes a new approach of an image descriptor based on Hilbert's curve, by using a keypoint statistical description. In terms of validation, the descriptor was compared to two other methods (BRISK and SIFT) using a classifier model. The results in the given conditions show a slightly better description capability than BRISK descriptor, using less memory but higher processing time spent.