Learning-based Approach for Occluded Face Recovery

• Main Authors: Mei-Chi Ho, 何美錡
• Other Authors: Ching-Ting Tu
• Format: Others
• Language: zh-TW
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/6kf3m5

碩士 === 淡江大學 === 資訊工程學系碩士班 === 103 === In this paper, we present a Bayesian framework for recovering the occluded facial image without the aid of manual face alignment. The proposed Bayesian framework unifies the recovery stage with the face alignment, and such complex probability distribution is solved represented by a particle set via a face prior. Into this framework, each particle is one possible pairwise solution of face alignment and recovery. First, the occluded facial patches of each particle are recovered by inferring their local facial details from other non-occluded patches. Further, by including the face prior knowledge as the constraint, the recovered results are robust to the local image noise which then cause the alignment parameters are accurately calculated. Particularly, we also propose a novel direct combined model (DCM)-based particle filter that utilizes the face specific prior knowledge to perform such particle-based solution efficiently and robustly. Our extensive experiment results demonstrate that the recovered images are quantitatively closer to the ground truth without manual involvement, and can be used for improving the accuracy of face reorganization application.