| Texto completo | |
| Autor(es): |
Costa‚ L.F.
Número total de Autores: 1
|
| Tipo de documento: | Artigo Científico |
| Fonte: | PATTERN RECOGNITION; v. 35, n. 11, p. 2445-2451, 2002. |
| Resumo | |
This article presents an effective spectral approach to estimate derivatives and curvature of open parametric curves. As the method is based on the discrete Fourier transform, the discontinuities of the curve (as well as of its derivatives) must be controlled to minimize the Gibbs phenomenon. We address this problem by obtaining a smooth extension of the curve in such a way as to suitably close it, which is done through a variational approach taking into account the spectral energy of differentiated versions of the extended curves. This novel method presents potential for applications in a broad class of problems, ranging from applied and experimental physics to image analysis. (C) 2002 Pattern Recognition Society. Published by Elsevier Science Ltd. All rights reserved. (AU) | |
| Processo FAPESP: | 99/12765-2 - Desenvolvimento e avaliação de métodos originais e precisos em análise de formas e imagens e visão computacional |
| Beneficiário: | Luciano da Fontoura Costa |
| Modalidade de apoio: | Auxílio à Pesquisa - Temático |