Descritor de características de imagem baseado na curva de preenchimento de espaço de hilbert aplicado em pontos de interesse

Abstract

Researchers have successfully applied Hilbert Space-Filling curve to improve many computing processes. By applying a space-filling curve approach focused on image keypoints, image description tasks could benefit from an alternative and potentially more efficient method. This work presents a color image feature descriptor based on the Hilbert space-filling curve applied to keypoints. Our approach utilizes the Scale Invariant Feature Transform (SIFT) just to detect keypoints from images. Subsequently, it employs a Hilbert curve to traverse a pixel neighborhood surrounding each detected keypoint for local descriptor calculation. From the values of the pixels within each Hilbert curve, we extract statistical measures and obtain a feature descriptor for each image using the Bag of Visual Words (BoVW) method. Our results showed an accuracy of 95%, surpassing the traditional BoVW and SIFT combination, demonstrating its potential for further improvement.