Parallel Dense Depth from Motion on the Image Understanding Architecture

This paper describes the design and implementation of a Single Instruction Multiple Data (SIMD) depth-from-motion algorithm on the Image Understanding Architecture Simulator. Correspondences are established in parallel for two temporally separated images through correlation. The correspondences are used to determine the translational and rotational motion parameters of the camera through a parallel motion algorithm. This is done by first determining the approximate translational parameters and then constraining the search for the exact translational and rotational parameters. Finally, the dense depth map is computed from the image correspondences and the computed motion parameters. Results are analyzed for three image sequences acquired from mobile vehicles (the Autonomous Land Vehicle, the Carnegie-Mellon NAVLAB, and the UMass Denning Robot). Depths are obtained at an average accuracy of about 8% in outdoor image sequences. The depth maps are processed to locate relatively small obsta..