Improved shape from shading using non-Lambertian reflectance models

This thesis focuses on the use of non-Lambertian reflectance models for curved surface analysis. The contributions are: a) to show how specular and diffuse reflectance models can be used to recover surface normal estimates using shape-from-shading (SFS), b) to show how wave scattering theory can be used to estimate rough surface characteristics, and c) to present a method for surface normal smoothing that imposes consistency on principal curvature directions. One of the problems that hinders standard methods for SFS is the presence of local specularities which may be misidentified as high curvature surface features. To address this problem we develop a maximum a posteriori probability estimation method for estimating the mixing proportions for Lambertian and specular reflectance, and also, for recovering local surface normals. The framework for our study is provided by the iterated conditional modes algorithm. Once improved estimates of surface normal directions are obtained, we can also remove the specular highlights from the raw image using the reconstructed specular intensity image. It is also well known that the non-specular reflectance from both rough and shiny surfaces depart significantly from the predictions of Lambert’s law. These effects are particularly marked at the occluding boundary where the surface is highly inclined to the viewer direction. Hence, Lambertian SFS methods can not be applied directly to the analysis of rough and shiny surfaces. In order to overcome this difficulty, we consider how to reconstruct the Lambertian component when the object is illuminated in the viewing direction. To do this we make use of the diffuse reflectance models described by Oren and Nayar, and by Wolff. Next, the thesis focuses on how reflectance models based on wave scattering theory and reported in the physics literature can be used for rough surface analysis. For dielectrics, the models investigated are derived from the Beckmann-Kirchhoff scatter theory, while for metals the starting point is the Davies scattering model. We show how these models can be used to estimate surface roughness parameters using simple reflectance