iVISE : interactive video search engine

• Main Author: Filzah Misran
• Other Authors: Hoi Chu Hong
• Format: Final Year Project
• Language: English
• Published: 2009
• Subjects: DRNTU::Engineering::Computer science and engineering::Information systems::Information storage and retrieval
• Online Access: http://hdl.handle.net/10356/19292
• Institution: Nanyang Technological University

This report serves to present the research and development process of the interactive Video Search Engine (iVISE) which is a content-based retrieval application that makes use of both keyword-based (text) and visual-based search. With the increase in popularity of digital devices, videos can now be easily captured and uploaded over the Internet. Despite this, most traditional search engines and video-sharing sites only employ simple keyword-based searches. Without understanding the video contents, these search engines may be limited in video retrieval. In order to understand the development of content-based retrieval, the report first reviewed the current technology used and the methods employed in available content-based retrieval prototype systems. By understanding the concept and procedures of content-based retrieval, iVISE is developed based on these methods. Next, the design period will draft out the essential modules of the application and describe what each module is and how they will function. Following that, the implementation period starts off with implementing the calling tool module which made use of a combination of YouTube-g and cclive. For the text search module, Ferret and Sphinx were compared and after testing out both text search libraries, Sphinx was chosen as it was able to outperform Ferret in terms of its indexing speed and also its search (retrieval) speed. Finally the visual search module was implemented with the use of FFmpeg, Features Extraction Library and also Fast Approximate Nearest Neighbor Search (FLANN) library.