Sabiia Seb
        Busca avançada

Botão Atualizar

Botão Atualizar

Registro completo
Provedor de dados:  InTech
País:  N/A
Título:  Automatic Facial Feature Extraction for Face Recognition
Autores:  Paola Campadelli
Raffaella Lanzarotti
Giuseppe Lipori
Data:  2007-07-01
Ano:  2007
Palavras-chave:  Face Recognition
Resumo:  The subject of this chapter is the presentation of a novel method for the automatic and precise localization of facial features in 2D still images. The method follows the top-down paradigm and consists of subsequent steps to decompose the initial problem in increasingly easier tasks: assuming a rough localization of the face in the image, first comes the application of an eye detector with the aim of discriminating between real face regions and possible false positives. The accuracy of the detection is nearly optimal. Successively, an eye locator is applied on a small neighborhood of the detector output to improve the localization precision. Finally, the eye center positions are used to derive 27 facial fiducial points, either extracted directly from the image or inferred on the basis of simple geometrical considerations. The eye localization module has been extensively tested on five publicly available databases with different characteristics to remark its generality. In the overall, the results are comparable to or better than those obtained by the state-of-the-art methods. The performance evaluation is carried out according to two objective performance measures in order to favor the comparison with other techniques. Concerning the fiducial point localization, results on the XM2VTS show high precision. In the last years many research works have pointed out the importance of facial feature localization as the fundamental step for the initialization of other methods, mostly face recognition techniques. In general, not all types of error affect the subsequent processing in the same way: for instance scale errors usually affect a FR technique more than translational
Tipo:  3
Idioma:  Inglês
Editor:  INTECH Open Access Publisher
Relação:  ISBN:978-3-902613-03-5

Empresa Brasileira de Pesquisa Agropecuária - Embrapa
Todos os direitos reservados, conforme Lei n° 9.610
Política de Privacidade
Área restrita

Parque Estação Biológica - PqEB s/n°
Brasília, DF - Brasil - CEP 70770-901
Fone: (61) 3448-4433 - Fax: (61) 3448-4890 / 3448-4891 SAC:

Valid HTML 4.01 Transitional