Image Cube Trajectory Analysis for Concentric Mosaics

We present a new concept for the extension of epipolar image analysis to more general camera configurations like circular camera movements usually occurring for concentric mosaic acquisition. In this way the robust method for 3D scene depth reconstruction which we call Image Cube Trajectory Analysis (ICT) is no longer restricted to horizontal, linear, and equidistant camera movements. Similar to epipolar image analysis the algorithm uses all available views of an image sequence simultaneously. Instead of searching for straight lines in the epipolar image we explicitly compute the trajectories of particular points through the image cube. Variation of the unknown depth leads to different curves. The best match is assumed to correspond to the true depth. It is selected by evaluating color constancy along the curve. For the example of concentric mosaics we derive an occlusion compatible ordering scheme that guarantees an optimal processing of occluding object points.