Estudo comparativo de técnicas de detecção de cantos em imagens digitais

Abstract

The corners are the most important points of the contour, they can represent efficiently a shape. The corners are characteristics often used in various tasks of Computer Vision, such as stereo vision, 3D rendering and motion estimation. This work presents analysis of performance and computational effort for different approaches to detection of corners existents in the literature using methods based on Wavelet Transform, Curvature and Mathematical Morphology, and proposes a new technique to detect corners, changing a Wavelet Transform technique. To perform the simulations Matlab / Simulink are used. The results enable to determine an algorithm that combines a high rate of detection of corners at a low computational effort.