Robust Matching By Dynamic Space Warping For Accurate Face Recognition.pdf

ROBUST MATCHING BY DYNAMIC SPACE WARPING FOR ACCURATE FACE RECOGNITION Hichem Sahbi and Nozha Boujemaa Imedia Research Group INRIA, BP 105, F-78153 Le Chesnay, France Email: Hichem.Sahbi,Nozha.Boujemaa
@inria.fr ABSTRACT framework to derive his CFR (Complex Feature Recogni- tion) algorithm. The utility of face recognition for multimedia indexing is Various methods for finding a body-centered frame have enhanced by using accurate detection and alignment of salient been suggested [6], including moments of inertia, symmetry invariant face features. The face recognition can be per- finders and a polar Fourier description, but these methods formed using template matching or a feature-based-approach, suffer from sampling and parameterisation errors and non- but both these methods suffer from occlusion and require uniqueness. To avoid these effects, Sclaroff and Pentland an a priori model for extracting information. To avoid these [7] developed the modal matching approach for correspon- drawbacks, we present in this paper a complete scheme for dence and recognition. They claim that the variation modes face recognition based on salient feature extraction in chal- provide a global-to-local ordering of shape deformations, lenging conditions, which is performed without an a priori which allow the selection of deformation types for align- or learned model. These features are used in a matching ment and comparison. Nastar and Pentland [8] described process that overcomes occlusion effects using the dynamic an approach for face matching which is the generalizati