Development of fuzzy manifold and fuzzy nearest distance for pose estimation of degraded face images

This paper presents a novel approach to estimate the pose position of the degraded face images using fuzzy vector representation. We focus on the development of fuzzy-manifold as an interpolation line between all the available learning data and the fuzzy distance calculation as the classifier. Crisp-based vector of the face image is firstly transformed into fuzzy-based vector, to deal with the fuzziness of the data caused by statistical measurement error directly. Two types of fuzzy interpolation methods are used to construct the fuzzy-manifold. Estimating the pose position of an unknown crisp-image vector is accomplished by firstly transformed into a fuzzy-vector and calculates the nearest fuzzy-distances to all available fuzzy-points in the designated fuzzy-lines. This system is then applied to estimate a degraded face images influenced by quality degradation effects. Experiment results show that those degradation effects due to various noise additions did not decrease the recognition rate.