A New Algorithm for Gradient Field Integration

Abstract

This paper proposes two improvements for reflectance based shape recovery. First, it is shown that albedo-independent photometric stereo allows albedo computation. This computation is based on photometric equations that relate surface normals to triplets of the image irradiances. Second, the paper also presents as the main result a new algorithm for depth recovery from surface normals. In order to improve the accuracy and robustness and to strengthen the relation between the depth map and surface normals, two new constraints are added into the associated cost function. They constrain the behavior of high-order change rate between the variables. Therefore, the changes of depth maps will be more regular. The Frankot-Chellappa-algorithm is a special case of our algorithm in the sense that it uses a subset of constraints only.