Ego-Motion Estimation and 3D Model Refinement in Scenes with Varying Illumination

We present an iterative algorithm for robustly estimating the ego-motion and refining and updating a coarse depth map using surface parallax and a generalized dynamic image (GDI) model. Given a coarse depth map acquired by a range-finder or extracted from a Digital Elevation Map (DEM), we first estimate the ego-motion by combining a global ego-motion constraint and a local GDI model. Using the estimated camera motion and the available depth estimate, motion of the 3D points is compensated. We utilize the fact that the resulting surface parallax field is an epipolar field and constrain its direction using the previous motion estimates. We then estimate the magnitude of the parallax field and the GDI model parameters locally and use them to refine the depth map estimates. We use a tensor based approach to formulate the depth refinement procedure as an eigen-value problem and obtain confidence measures for determining the accuracy of the estimated depth values. These confidence measures are used to remove regions with potentially incorrect depth estimates for robustly estimating ego-motion in the next iteration. Experimental results using both synthetic and real data are presented. Comparisons with results obtained using a brightness constancy (BC) model show that the proposed algorithm works significantly better when time-varying illumination changes are present in the scene.