Shape from shading using the facet model

摘要

In this paper, the facet model is used in the recovery of surface orientation from single and multiple images. Two methods for determining surface orientation from image shading are presented. The first method works in the single image domain and is formulated as a non-linear optimization problem. Since three-dimensional scene information available in a single image is usually ambiguous, the optimization procedure can result in multiple solutions. The possibility of adding boundary constraints to the optimization process is also investigated. In the second method, additional images are obtained from the same viewing position, but with changing illumination direction. With these additional images, local surface orientation is determined uniquely by a linear optimization procedure. Experimental results of the shape-from-shading methods are also described.