Abstract
In this paper we consider the problem of recovering depth from a moving camera with an attached light. The fact that the light is moving means that the appearance of the scene changes, making intensity-based correlation difficult. These appearance changes, however, are governed by the motion of the camera, and are intimately related to the depth of the scene. This allows us to derive constraints on the intensities of corresponding points. We show that, in principle, this approach allows us to recover depth and surface albedo (up to a scale from light intensity) from two calibrated views. This is possible even when viewing textureless surfaces, which are intractable for traditional stereo methods. However, the method is very sensitive to quantisation of intensity values, and so is not practical on its own for normal camera images. The underlying constraints, however, may be useful as part of larger systems.