Visual Concept Detection in Images and Videos

The rapidly increasing proliferation of digital images and videos leads to a situation where content-based search in multimedia databases becomes more and more important. A prerequisite for effective image and video search is to analyze and index media content automatically. Current approaches in th...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
1. Verfasser: Mühling, Markus
Beteiligte: Freisleben, Bernd (Prof. Dr.) (BetreuerIn (Doktorarbeit))
Format: Dissertation
Sprache:Englisch
Veröffentlicht: Philipps-Universität Marburg 2014
Schlagworte:
Online Zugang:PDF-Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:The rapidly increasing proliferation of digital images and videos leads to a situation where content-based search in multimedia databases becomes more and more important. A prerequisite for effective image and video search is to analyze and index media content automatically. Current approaches in the field of image and video retrieval focus on semantic concepts serving as an intermediate description to bridge the “semantic gap” between the data representation and the human interpretation. Due to the large complexity and variability in the appearance of visual concepts, the detection of arbitrary concepts represents a very challenging task. In this thesis, the following aspects of visual concept detection systems are addressed: First, enhanced local descriptors for mid-level feature coding are presented. Based on the observation that scale-invariant feature transform (SIFT) descriptors with different spatial extents yield large performance differences, a novel concept detection system is proposed that combines feature representations for different spatial extents using multiple kernel learning (MKL). A multi-modal video concept detection system is presented that relies on Bag-of-Words representations for visual and in particular for audio features. Furthermore, a method for the SIFT-based integration of color information, called color moment SIFT, is introduced. Comparative experimental results demonstrate the superior performance of the proposed systems on the Mediamill and on the VOC Challenge. Second, an approach is presented that systematically utilizes results of object detectors. Novel object-based features are generated based on object detection results using different pooling strategies. For videos, detection results are assembled to object sequences and a shot-based confidence score as well as further features, such as position, frame coverage or movement, are computed for each object class. These features are used as additional input for the support vector machine (SVM)-based concept classifiers. Thus, other related concepts can also profit from object-based features. Extensive experiments on the Mediamill, VOC and TRECVid Challenge show significant improvements in terms of retrieval performance not only for the object classes, but also in particular for a large number of indirectly related concepts. Moreover, it has been demonstrated that a few object-based features are beneficial for a large number of concept classes. On the VOC Challenge, the additional use of object-based features led to a superior performance for the image classification task of 63.8% mean average precision (AP). Furthermore, the generalization capabilities of concept models are investigated. It is shown that different source and target domains lead to a severe loss in concept detection performance. In these cross-domain settings, object-based features achieve a significant performance improvement. Since it is inefficient to run a large number of single-class object detectors, it is additionally demonstrated how a concurrent multi-class object detection system can be constructed to speed up the detection of many object classes in images. Third, a novel, purely web-supervised learning approach for modeling heterogeneous concept classes in images is proposed. Tags and annotations of multimedia data in the WWW are rich sources of information that can be employed for learning visual concepts. The presented approach is aimed at continuous long-term learning of appearance models and improving these models periodically. For this purpose, several components have been developed: a crawling component, a multi-modal clustering component for spam detection and subclass identification, a novel learning component, called “random savanna”, a validation component, an updating component, and a scalability manager. Only a single word describing the visual concept is required to initiate the learning process. Experimental results demonstrate the capabilities of the individual components. Finally, a generic concept detection system is applied to support interdisciplinary research efforts in the field of psychology and media science. The psychological research question addressed in the field of behavioral sciences is, whether and how playing violent content in computer games may induce aggression. Therefore, novel semantic concepts most notably “violence” are detected in computer game videos to gain insights into the interrelationship of violent game events and the brain activity of a player. Experimental results demonstrate the excellent performance of the proposed automatic concept detection approach for such interdisciplinary research.
DOI:10.17192/z2014.0483