Robust and Efficient 3-d Optical Flow Computing Framework

This book presents a framework that improves the robustness and accuracy in computing dense optical-flow fields. We propose a global formulation with a regularization term that can handle the 3D flow patterns, such as rotation, zoom, and their combinations, induced by a 3D rigid-body motion. The movitation of this book is that the traditional optical flow regularizers incorrectly penalize the above 3D flow patterns and result in biased estimates. Hence, the purpose of this work is to derive a new suite of regularization expressions that treat all valid flow patterns resulting from a 3D rigid-body motion equally, without unfairly penalizing any of them. Furthermore, we develop a framework of optical flow computation and flow model inference method. The proposed framework can fully utilize the power of the proposed 3D rigid body motion regularizers. The proposed framework is well suited for implementing real-time and high accuracy image and video processing applications.