2015 | OriginalPaper | Buchkapitel
An Investigation of fSVD and Ridgelet Transform for Illumination and Expression Invariant Face Recognition
verfasst von : Belavadi Bhaskar, K. Mahantesh, G. P. Geetha
Erschienen in: Advances in Intelligent Informatics
Aktivieren Sie unsere intelligente Suche, um passende Fachinhalte oder Patente zu finden.
Wählen Sie Textabschnitte aus um mit Künstlicher Intelligenz passenden Patente zu finden. powered by
Markieren Sie Textabschnitte, um KI-gestützt weitere passende Inhalte zu finden. powered by
This paper presents a wide-eyed yet effective framework for face recognition based on the combination of flustered SVD(fSVD) and Ridgelet transform. To this end we meliorate in the sense of computation efficiency, invariant to facial expression and illumination of [21]. Firstly fSVD is applied to an image by modelling SVD and selecting a proportion of modelled coefficients to educe illumination invariant image. Further, Ridgelet is employed to extract discriminative features exhibiting linear properties at different orientations by representing smoothness along the edges of flustered image and also to map line singularities into point singularities, which improves the low frequency information that is useful in face recognition. PCA is used to project higher dimension feature vector onto a low dimension feature space to increase numerical stability. Finally, for classification five different similarity measures are used to obtain an average correctness rate. We have demonstrated our proposed technique on widely used ORL dataset and achieved high recognition rate in comparision with several state of the art techniques.