Multiresolution image acquisition and surface reconstruction using active stereo

Abstract

This paper is concerned with the problem of surface reconstruction from stereo images for large scenes having large depth ranges. The passive stereo paradigm is inadequate for this problem because of the need to aim cameras in different directions and to fixate at different objects. We present an active stereo approach in which the scene is systematically scanned and image acquisition and surface reconstruction are integrated using a four-step process. First, a new fixation point is selected from among the nonfixated, low resolution scene parts of current fixation. Second, a reconfiguration of cameras is initiated for refixation. As reconfiguration progresses, the images of the new fixation point gradually deblur and the accuracy of the stereo estimate of the point improves. In the third step, the improved stereo estimate is used to achieve accurate focus and vergence settings of the cameras for fixation. Finally, focus-based depth estimates are obtained at a grid near the fixation point whose density is determined by the local surface slope. These estimates are fused with those obtained from stereo using maximum likelihood (weighted averaging) and are interpolated to the non-grid points.