The purpose of this study is to investigate a new method for recovering relative depth and 3D motion from a temporal sequence of monocular images. The method is direct insomuch as it does not require computation of image motion prior to 3D interpretation. This interpretation is obtained by minimizing a functional with two characteristic terms, one of conformity to the spatiotemporal changes in the image sequence, the other of regularization based on anisotropic diffusion. The Euler-Lagrange equations corresponding to the functional minimization are solved iteratively via the half-quadratic algorithm.
Citation:
H. Sekkati, A. Mitiche, "Dense 3D Interpretation of Image Sequences: A Variational Approach Using Anisotropic Diffusion," iciap, pp.424, 12th International Conference on Image Analysis and Processing (ICIAP'03), 2003