Methodology for Automatic Segmentation, Indexing And Content-based Retrieval of Presentation Video

INVESTIGATORS

Dr. Horace H S Ip, Mr. Chan Siu Lok

 

BRIEF DESCRIPTION

An innovative methodology for automatic segmentation, indexing and content-based retrieval of digital video is introduced. This methodology is specially-designed for manipulating presentation videos such as seminars or university lectures. To allow non-sequential searching and browsing, lecture video should be partitioned and indexed according to its content. Utilizing the properties of most lecture videos, video boundaries could be detected automatically according to slide changes and video-shots could be partitioned without using any pre-defined image models. The content of the video-shot would be extracted from its representation frame. With the aid of the original text materials, hierarchical video indexes based on the logical structure of lecture text could be built. Students would be able to retrieve the lecture in different levels of topics. The indexed video-shots could be used to construct educational content-based retrieval system or CAI systems. 

We have designed and implemented two modules for this system. The video segmentation/indexing module is used to parse, extract content and build hierarchical index for lecture video as shown in Figure 1. The video browsing/query module is used to browse and search for lecture video according to specific query condition. This module was constructed dynamically by former module according to the video parsing and indexing result. HTML and JavaScript code would be generated and displayed on WWW browsers. Figure 2 shows the user-interface of this module in MS Internet Explorer <http://cpcs78/lvsegm.htm>. On this interface, the left side of the screen shows the hierarchical index of a lecture video talking about UNIX. On clicking such index node, it would link to an associated video shot which is displayed on the lower part of the right side. The upper part shows the original lecture text. Personal annotations and content search of video are supported. This module is designed for various kinds of educational video for the purpose of information repository and distance learning on Internet.

FUNDING AGENCY

City University of Hong Kong

 

PUBLICATIONS

  1. Chan, S.L., Ip, H.H.S. "Efficient Methodology for Automatic Video-Content Indexing and Retrieval of Lectures for CAI Systems", Proceeding of ED-MEDIA 96, pp137-pp141, Boston, USA, June 1996.
  2. Ip, H.H.S, Chan, S.L. (accepted) "Hypertext-Assisted Video Indexing and Content-based Retrieval", 8th ACM International Hypertext Conference , Southampton, British, April 1997.
  3. Ip, H.H.S., Chan, S.L., (to be appeared) "Automatic Segmentation and Index Construction for Lecture Video", Journal of Educational Multimedia & Hypermedia, USA, 1997.

 

FIGURES OR IMAGES

(60kb)

Figure 1 : The video segmentation/indexing module of the system

(97.3kb)

Figure 2 :  Video Browsing/Query Module on MS Internet Explorer