We investigate the recovery of 3-D motion and structure from the stereo images of a stationary environment. A Kalman filter-based framework is proposed for the reconstruction of 3-D structure from multiple visual cues, through the integration of image motion and stereo disparity with the shading flow that is induced by the rotational motion of the source. A common scenario involves the coupled motion of artificial source(s) and stereo cameras that are installed on mobile submersible vehicles. Utilizing shading flow with the image motion leads to devising a more robust 3-D motion estimation algorithm, in addition to the important role in depth recovery/refinement by constraining the local surface gradients. Collectively, use of multiple cues enhances robustness with respect to perturbation in any of the cues. Results of experiments with synthetic and real imagery are presented to evaluate the performance of the proposed algorithm, including the construction of a composite 3-D depth map of an underwater scene from a sequence of stereo pairs.
- 3-D reconstruction
- Image mosaicking
- Motion estimation
ASJC Scopus subject areas
- Computer Vision and Pattern Recognition
- Signal Processing
- Electrical and Electronic Engineering