A Markov Random Field Groupwise Registration Framework for Face Recognition
Complexities in the facial recognition increases because of real time image acquisition, which is generally not performed by an expert. Because of the inappropriate focus, the positional aspect of the input image can be different. In this work, a directional aspect based structural analysis is provided for generating the Local binary pattern. For a single face about 60 different binary patterns are obtained in the positional variation of 30 degrees. As the structural pattern set is obtained, the HMM-PCA model is applied for facial mapping. The segmented HMM model is applied to probabilistic facial segment map which is followed by ageing face mapping using PCA approach. The work is applied to Aberdeen Color Face Dataset, Iranian dataset and Yale Face Datasets. The comparative results are obtained against the PCA and LDA approaches which show that the presented work is significantly better with higher accuracy.