Interactive video retrieval

Huang, Zheng (2005) Interactive video retrieval. MSc(R) thesis, University of Glasgow.

Abstract

Video storage, analysis, and retrieval has become an important research topic recently due to the advancements in the creation and distribution of video data. In this thesis, an investigation into interactive video retrieval is presented. Advanced feedback techniques have been investigated in the retrieval of textual data. Novel interactive schemes, mainly based on the concept of relevance feedback, have been developed and experimented. However, such approaches have not been applied in the video retrieval domain. In this thesis, we investigate the use of advanced interactive retrieval schemes for the retrieval of video data. To understand the role of various features for the video retrieval, we experimented with various retrieval strategies. We benchmarked the role of visual features, the textual features and their combination. To explore this further, we categorized query into various classes and investigated the retrieval effectiveness of various features and their combination. Based on the results, we developed a retrieval scheme for video retrieval. We developed an interactive retrieval technique based on the concept of implicit feedback. A number of retrieval models are developed based on this concept and benchmarked with a simulation- based evaluation strategy. A Binary Voting Model performed well and has been reformed for user-based experiments. We experimented with the users and compared the performance of an interactive retrieval system, using a combination of implicit and explicit feedback techniques, with that of a system using explicit feedback techniques.

• Item Type: Thesis (MSc(R))
• Qualification Level: Masters
• Keywords: Computer science.
• Colleges/Schools: College of Science and Engineering > School of Computing Science
• Supervisor's Name: Jose, Dr. Joeman
• Date of Award: 2005
• Depositing User: Enlighten Team
• Unique ID: glathesis:2005-71097
• Copyright: Copyright of this thesis is held by the author.
• Date Deposited: 10 May 2019 10:49
• Last Modified: 10 Sep 2021 09:27
• Thesis DOI: 10.5525/gla.thesis.71097
• URI: https://theses.gla.ac.uk/id/eprint/71097