Aristotle University Of Thessaloniki (AUTH)
Σημασιολογική ανάλυση πολυμεσικού περιεχομένου με χρήση τεχνικών μηχανικής μάθησης
Abstract
dc:descriptionIn this thesis, the problem of the semantic analysis of the multimedia content using machine learning techniques, and in particular the issue of detecting high-level semantic concepts in static images and video sequences, is studied. Specifically, an approach to semantic image analysis and classification is initially proposed that is based on the combination of global and local image features, making use of domain-specific prior knowledge. Subsequently, a spatial context exploitation technique is presented, which aims at improving the image analysis results that have been computed using only visual information. Additionally, an extensive comparative evaluation of the proposed method with other similar techniques of the literature is conducted. The aim of this study is the in-depth investigation of the advantages of each technique and the gain of a better insight on the use of spatial context. Moreover, a novel approach to Hidden Markov Model (HMM)-based semantic video analysis, which relies on the statistical analysis and representation of the motion signal at local level, is presented. In parallel, a variant of the proposed approach for providing motion information to HMMs is also introduced, which presents significantly reduced computational complexity, while maintaining a recognition performance similar to that of the original method. Finally, an approach to semantic video analysis is proposed for jointly performing modality fusion and temporal context exploitation.
Degree
thesis:*- Grantor dc:publisher
- Aristotle University Of Thessaloniki (AUTH)
- Year dc:date
- 2011
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Παπαδόπουλος, Γεώργιος
- Papadopoulos, Georgios
Subjects
dc:subject × 12- Σημασιολογική ανάλυση πολυμεσικού περιεχομένου
- Μηχανική μάθηση
- Αξιοποίηση πληροφορίας πλαισίου
- Ανάλυση κίνησης
- Semantic multimedia content analysis
- Machine learning
- Contextual information exploitation
- Motion analysis
- Επιστήμες Μηχανικού και Τεχνολογία
- Επιστήμη Ηλεκτρολόγου Μηχανικού, Ηλεκτρονικού Μηχανικού, Μηχανικού Η/Υ
- Engineering and Technology
- Electrical Engineering, Electronic Engineering, Information Engineering
Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/23953