Hair Modeling using Kinect Sensors and DSLR Cameras

Abstract

3D Hair reconstruction based on real-life hair capture is an important and challenging work in hair modeling field. Most of existing hair capture methods use 2D images to reconstruct 3D hair, and these methods usually adopt 3D polygons to present hair wisp. In this paper, we introduce an approach to capture real-life hair using Kinect sensor and digital single-lens reflex (DSLR) camera and to reconstruct 3D hair model using particle system. First, our method collects four views of point clouds and high resolution image for real-life hair. We register DSLR image and point cloud to build the mapping relationship between 2D and 3D and the alignment techniques are utilized to merge the point clouds. With the manually extracted 2D hair strands from the DSLR image, the system used control points to represent hair strands as spline curve. Furthermore, these control points are projected on the point cloud to find the corresponding 3D control points. Finally the system reconstructs 3D hair model where the strands are represented in particle system. We also present a hidden hair pieces recovery algorithm to generate final well-connected 3D hair strands. Our method is novel and has many advantages: (i) hardware setting is simple and affordable (ii) combination of high quality image of DSLR and depth of Kinect taking advantage of each of them (iii) the 2D and 3D combined method allows us to repair and improve the quality of 3D depth (iv) Hair representation is spline based which is a particle system and most common hair animation is based on particle system.