High-Quality Single Image 3D Facial Shape Reconstruction via Robust Albedo Estimation

In this paper, we present a method for high-quality single image 3D facial shape reconstruction with robust albedo estimation by using a prior face model. Motivated by the state-of-the-art methods for single image-based face reconstruction using the Shape from Shading (SfS), we develop a 3D facial recovery framework with innovative components. Specifically, rather than using the mean albedo, we use an optimized albedo of the 3D Morphable Model (3DMM) as a prior albedo model. In addition, we develop a robust weighting method based on the chromaticity of input and back-projected images. Combining these approaches produces a more robust albedo estimation which directly enhances the result of the SfS process. By using the tetrahedron-based deformation technique, a detailed facial shape can be accurately recovered from a single image. We demonstrate that our 3D facial recovery framework can reconstruct fine geometric details for a single image and outperforms the state-of-the-art in single image face reconstruction.