Combining Block-based PCA, Global PCA and LDA for Feature Extraction in Face Recognition

Denis H. P. SalvadeoDébora C. CorrêaAlexandre L. M. LevadaNelson D. A. MascarenhasJander MoreiraJosé H. Saito

This paper presents a novel methodology for pattern classification, combining different feature extraction approaches, more precisely, global PCA, block-based PCA and LDA in order to improve the face classification performance. Block-based PCA is a new technique that is being widely used for feature extraction, especially in face recognition. Basically, the idea of block-based PCA is to first divide each image into several kk × blocks and then perform a dimensionality reduction, mapping a block of pixels into one single pixel. Experiments with face images from the ORL database present good results, showing that the proposed combination strategy for feature extraction is suitable for face recognition problems.

Caso o link acima esteja inválido, faça uma busca pelo texto completo na Web: Buscar na Web

Biblioteca Digital Brasileira de Computação - Contato:
     Mantida por: