Intelligent video segmentation for extracting video objects

• Main Author: Mao, Jinghong.
• Other Authors: Ma, Kai-Kuang
• Format: Theses and Dissertations
• Published: 2008
• Subjects: DRNTU::Engineering::Electrical and electronic engineering::Electronic systems::Signal processing
• Online Access: http://hdl.handle.net/10356/4453
• Institution: Nanyang Technological University

In order to meet the demand of modern multimedia technology, which shows an increasing interest in content-based manipulation of video information, Video object (VO) is introduced in MPEG-4 to address content-based functionalities. Therefore, an effective VO segmentation is a crucial processing step in modern digital video processing. However, VO segmentation is intrinsically an ill-posed challenge and encounters three major concerns: computational complexity, in-accurate boundaries of segmented VOs, and integration of user's interaction. Although motion segmentation and spatio-temporal segmentation have their individual applications and advantages, the trend of the modern video process-ing methodology not only focus on the low-level features such as intensity/color, motion but also introduces the high-level semantic information to bridge the gap between the human visual system and computer processing. In this the-sis, we investigate three methodologies—motion segmentation, spatio-temporal segmentation, and semantic segmentation, for VO segmentation and contribute our new solutions to handle above-mentioned issues.