Global Motion Estimation Based on the Constrained Spatio-temporal Gradient Method in Model-Based Coding

For global motion estimation in model-based coding, this paper proposes a constrained spatio-temporal gradient method using contour information. To overcome the local minimum problem in motion tracking of the conventional spatio-temporal gradient method, the translational matching position detected based on 2D contour matching and the rotational motion model are introduced as constraints. Simulation results show that the proposed method yields smaller mean square error (MSE)than the conventional method.

• Type: Journal
• Year: 2002
• Where: MVA
• Authors: Young Wook Sohn, Doo-Hyun Kim, Dong-O Kim, Rae-Hong Park
• Keywords: Computer Vision, Global Motion Estimation, MVA 2002, Spatio-temporal Gradient Method, Translational Matching Position