New Shape-From-Shading Method With Specular Reflectance Model

Lei Yang,Weimin Ma

2009-01-01

Abstract:Shape-from-shading(SFS) is an approach of three dimension shape reconstruction from only one grey image in computer vision. Reflectance models and reconstruction algorithms are two key factors of SFS. Existing SFS methods are mostly based on the simple Lambertian reflectance model. But real surfaces such as metal show greatly specular reflectance. New SFS method based on specular reflectance model is proposed in this paper. Firstly, image reflectance map equation is formulated based on the specular Phong model. Then the reflectance map equation turns into a static Hamilton-Jacobi equation named Eikonal equation. And the SFS problem is formulated as viscosity solution of the static Hamilton-Jacobi equation. Following the fast marching algorithm is used to solve the Hamilton-Jacobi equation and new SFS method is achieved. At last, experiments on both synthetic and real images are performed. Experimental results show that the proposed SFS method are more accurate than Lambertian reflectance-based SFS method under the same condition when dealing with real surface of specular reflectance.