Manifold Estimation in View-based Feature Space for Face Synthesis Across Pose

This paper presents a new approach to synthesize face images under different pose changes from a single input image. The approach is based on two observations: 1. face images from a single person under different poses could be mapped to a smooth manifold in a unified feature space. 2. the manifolds from different faces are separated from each other by their similarities. The new manifold estimation is formulated as a minimization problem with smoothness constraints. The experiments show that face images under different poses can be robustly synthesized from one input image, even with large pose changes.