Technique for 3D Face Reconstruction from Depth Images

Abstract

Here we are giving a novel framework for ThreeD face regetting from an single depth image is proposed. We make use of an inexpensive Kinect® sensor which provides a depth map of the subject. The depth map obtained is of low resolution having substantial random noise and corruption which makes the 3D face reconstruction problem more difficult to implement. We introduce a novel algorithm that effectively maps the low resolution depth maps to realistic face models. We extract the sparse errors from the depth images using a data-driven local sparse coding technique and reconstruct the entire shape by combining local shapes using a template based method. Our approach is able to produce high-resolution 3D face models with better accuracy.