Probabilistic Elastic Matching based Face Verification under Pose Variant
Zhang Ling-l · Shuxue de shijian yu renshi · 2015
In order to solve real world face verification under pose variation,a probabilistic elastic matching approach is proposed.This method firstly extracts local features from densely sampled multi-scale image patches.By augmenting each feature with its location,a Gaussian mixture model is trained to capture the spatial-appearance distribution of all the training face images.Each Gaussian component builds correspondence of a pair of features to be matched between two faces.In process of face verification,a Support Vector Machine classifier is trained on the vector concatenating the difference vectors of all the feature pairs to decide if a pair of faces is matched or not.Experimental results show that the proposed method outperforms the state-of-the-art on common image database.