Object Segmentation and Watermarking for Video Coding

碩士 === 國立臺灣大學 === 電信工程學研究所 === 90 ===  MPEG-4 is a content-based video coding standard. It focuses on offering video storage and transmission over the channel at very low bit-rate. In MPEG-4 standard, the video is composed of many objects. Before transmitted, each object is coded through...

Full description

Bibliographic Details
Main Authors: Chang, Chia Pei, 張嘉珮
Other Authors: 貝蘇章
Format: Others
Language:en_US
Published: 2002
Online Access:http://ndltd.ncl.edu.tw/handle/71504314442827842540
Description
Summary:碩士 === 國立臺灣大學 === 電信工程學研究所 === 90 ===  MPEG-4 is a content-based video coding standard. It focuses on offering video storage and transmission over the channel at very low bit-rate. In MPEG-4 standard, the video is composed of many objects. Before transmitted, each object is coded through texture coding and shape coding individually. Users can synthesize a new video sequence with different objects at the receiver end.  In this thesis, we develop a segmentation system and watermarking techniques for video objects. The proposed segmentation system is based on change detection. With some simple morphological filters and the statistical background information, the moving objects can be segmented out of static background efficiently. The texture watermark of video objects is embedded in the bit-planes of the image using visual threshold scheme. As to the shape watermark, we propose a new technique that modifies the object contour slightly and embeds the watermark in the statistical properties of the contour. Experimental results show the practicability and robustness of the proposed methods.