الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 140 |  |  |
| | | from | 140 | | |
Abstract
The vast growth and progress of multimedia and videos on the Internet caused a growing demand for video retrieval, and search engines, as many users require content-based retrieval engines over text-based retrieval, misusing copyrighted video materials by bootleg manipulated videos is also a problem that requires detecting before uploading. This thesis introduces an effective implementation for a content-based video search engine using a large-scale dataset based on YouTube video links, which contain 1088 videos, that represent more than 65 hours of video, 11,000 video shots, and 66,000 unmarked and marked keyframes, 80 different objects names used for marking, with 58 objects categories of video records are indexed and classified in the video index dataset. Moreover, an effective features vector is introduced using video shots’ temporal and key-objects/concepts features by applying combinational-based matching algorithms, using various similarity metrics for evaluation. The retrieval system was evaluated using more than 200 non-semantic-based video queries evaluating both normal and bootleg videos, with retrieval precision for normal videos of 97.9% and retrieval recall of 100% combined by the F1 measure to be 98.3%. Bootleg video retrieval precision scored 99.2% and retrieval recall of 96.7% combined with the F1 measure to be 97.9%. This concluded that this technique would help in both enhancing traditional text-based search engines and promoting commonly used bootleg video detection techniques.