Cheng, Xin (1999) Feature-based motion estimation and motion segmentation. Masters thesis, Memorial University of Newfoundland.
PDF (Migrated (PDF/A Conversion) from original format: (application/pdf))
- Accepted Version
Available under License - The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission.
With the advancement of multimedia and hypermedia, video sequence segmentation is becoming more and more important. This thesis presents a new paradigm in the field of video sequence segmentation research. This new approach employs a limited number of landmarks to represent image features and segments video sequences into distinct moving objects according to the relation between different kinds of motion. The strength of this approach lies in the fact that the motion computations are performed only on the landmarks. By adjusting the number and positions of the landmarks, the scheme controls the motion estimation and motion segmentation procedure. -- To serve distinct purposes, the scheme uses three different approaches to produce landmarks: the manual mode, the tiling mode, and the automatic mode. For these three modes, the selection of landmarks, motion estimation, clustering, and motion segmentation are applied on standard video sequences. Two new methods for identifying landmarks automatically have been developed and are presented. These methods are compared with typical feature point detectors over different kinds of image pictures, showing that the new methods are more efficient than previous methods. One novel aspect of the scheme is the coupling of automatic landmark identification with motion estimation, producing very accurate motion estimation. -- As an application, bubble tracking for multiphase fluid flow is presented and demonstrated. To validate the scheme, a motion analysis platform has been developed to implement the overall scheme. This platform, with comprehensive functions for motion analysis and image processing, makes a practical contribution to video analysis.
|Item Type:||Thesis (Masters)|
|Additional Information:||Bibliography: leaves 95-97|
|Department(s):||Engineering and Applied Science, Faculty of|
|Library of Congress Subject Heading:||Image analysis; Image processing|
Actions (login required)