Parallelization between face localization and person identification

The face localization is the most important process of automatic facial recognition. If facial position is estimated by using a geometrical model, the estimation accuracy deteriorates unless the model is optimal for facial appearance. In this work, we deal with variations of facial scale and individuals, and we propose an efficient method that estimates facial position in parallel with facial scale and a personal model. The process based on the dynamic link architecture and the scale conversion iteratively estimates facial position and scale. The proposed method estimates the personal model by using a hierarchical face model database. As facial position estimation proceeds, the method selects the optimal model according to the hierarchical structure of the database. We can regard the selected personal model as a person identification result. We implement this facial recognition system based on the proposed method and demonstrate the advantages of the system through some experiments.