Most of the automated video-surveillance applications are based on background (BG) subtraction techniques, that aim at distinguishing moving objects in a static scene. These strategies strongly depend on the BG model, that has to be initialized and updated. A good initialization is crucial for the successive processing. In this paper, we propose a novel method for BG initialization and recovery, that merges interesting ideas coming from the video inpainting and the generative modelling subfields. The method takes as input a video sequence, in which several objects move in front of a stationary BG. Then, a statistical representation of the BG is iteratively built, discarding automatically the moving objects. The method is based on the following hypotheses: (i) a portion of the BG, called sure BG, can be identified with high certainty by using only per-pixel reasoning and (ii) the remaining scene BG can be generated utilizing exemplars of the sure BG. The proposed algorithm is able to exploit these hypotheses in a principled and effective way.

To my family and friends ii This thesis investigates semantic analysis of broadcast sports footage. A domain depen-dent sports video model is proposed. Under this model, the game semantics can be derived according to their relationship with the sequence of dynamic events that occur in the sport and the evolution of the spatio-temporal behaviour of a relevant object. Snooker and tennis are targeted as typical broadcast sports footage for the purpose of this research. The prob-lem focus is to automatically extract semantically meaningful events and to convey a useful representation to the user. Access to semantics provides a more natural tool for a user to query a corpus of data than by low-level content based features alone. These semantics are however open to various interpretations by different viewers. Therefore, in order to create a successful semantic based retrieval system it is necessary to consider the user-context. Unconstrained sports footage is generally very complicated in structure, so restricting the domain being addressed enables

In this report the development of a specialised search engine for a digital library is described. The proposed system architecture consists of three levels: the conceptual, the logical and the physical level. The conceptual level schema enables by its exposure of a domain specific schema semantically rich conceptual search. The logical level provides a description language to achieve a high degree of flexibility for multimedia retrieval. The physical level takes care of scalable and e#cient persistent data storage. The role, played by each level, changes during the various stages of a search engine's lifecycle: (1) modeling the index, (2) populating and maintaining the index and (3) querying the index. The integration of all this functionality allows the combination of both conceptual and content-based querying in the query stage. A search engine for the Australian Open tennis tournament website is used as a running example, which shows the power of the complete architecture and its various components.

Since the invention of television, and later the Internet, the amount of video content available has been growing rapidly. The great mass of visual material is an invaluable source of information, but its usefulness is limited by the available means of accessing and tailoring it to the needs of an individual. Long experience with text as a medium of conveying information allowed us to develop relatively effective methods of dealing with textual data. Unfortunately, the currently available techniques of accessing and processing video data are largely inadequate to the needs of its potential users. Hence video material remains a valuable but grossly untapped resource. In the domain of video news sources, this problem is especially severe. Television news stations broadcast continuous up-to-the-minute information from around the globe. For any individual viewer, only small portions of this news stream is of interest, yet currently no methods exist which would allow him to filter and monitor only the interesting news.

As amounts of publicly available video data grow, the need to automatically infer semantics from raw video data becomes significant. In this paper, we focus on the use of Dynamic Bayesian Networks (DBNs) for that purpose, and demonstrate how they can be effectively applied for fusing the evidence obtained from different media information sources. The approach is validated in the particular domain of Formula 1 race videos. For that specific domain we introduce a robust audiovisual feature extraction scheme and a text recognition and detection method. Based on numerous experiments performed with DBNs, we give some recommendations with respect to the modeling of temporal and atemporal dependences within the network. Finally, we present the experimental results for the detection of excited speech and the extraction of highlights, as well as the advantageous query capabilities of our system.

this paper, we present a content-based video retrieval system called Cobra. The system supports automatic extraction and retrieval of high-level concepts (such as video objects and events) from raw video data. It benefits from using domain knowledge, but at the same time, provides a general framework that can be used in different domains

Content-based retrieval has been identified as one of the most challenging problems, requiring a multidisciplinary research among computer vision, information retrieval, artificial intelligence, database, and other fields. In this paper, we address the specific aspect of inferring semantics automatically from raw video data. In particular, we present the Cobra video database management system that supports the integrated use of different knowledge-based methods for mapping low-level features to high-level concepts. We focus on dynamic Bayesian networks and demonstrate how they can be effectively used for fusing the evidence obtained from different media information sources. The approach is validated in the particular domain of Formula I race videos. For that specific domain we introduce a robust audio-visual feature extraction scheme and a text recognition and detection method. Based on numerous experiments performed with DBNs, we give some recommendations with respect to the modeling of temporal dependences and different learning algorithms. Finally, we present the experimental results for the detection of excited speech and the extraction of highlights, as well as the advantageous query capabilities of our system.

While it is very hard to achieve automatic sports competition key moments detection only based on visual analysis, we propose in this paper automatic highlights detection based on an audio classifier. The audio classifier is based on a new modeling technique of the audio spectrum called Piecewise Gaussian Modeling (PGM) and Neural Networks. The proposed approach was evaluated on soccer and tennis videos, though our technique has no restriction on the sports ’ type. It is shown that audio-based highlights detection can be effective for tennis segmentation since 97.5 % of end-of-serves were correctly classified. Goals can be detected in soccer videos using audio analysis as well. An intelligent sports-videos player is proposed based on the audio analysis permitting the user to navigate through key moments in a sports video. 1.

Most of the automated video-surveillance applications are based on the process of background modelling, aimed at discriminating motion patterns of interest at pixel, region or frame level in a nearly static scene. The issues characterizing an ordinary background modelling process are typically three: the background model representation, the initialization, and the adaptation. This paper proposes a novel initialization algorithm, able to bootstrap an integrated pixeland region-based background modelling algorithm. The input is an uncontrolled video sequence in which moving objects are present, the output is a pixel- and region-level statistical background model describing the static information of a scene. At the pixel level, multiple hypotheses of the background values are generated by modelling the intensity of each pixel with a Hidden Markov Model (HMM), also capturing the sequentiality of the different color (or gray-level) intensities. At the region level, the resulting HMMs are clustered with a novel similarity measure, able to remove moving objects from a sequence, and obtaining a segmented image of the observed scene, in which each region is characterized by a similar spatio-temporal evolution. Experimental trials on synthetic and real sequences have shown the effectiveness of the proposed approach.

Most of the automated video-surveillance applications are based on the process of background modelling, aimed at discriminating motion patterns of interest at pixel, region or frame level in a nearly static scene. The issues characterizing an ordinary background modelling process are typically three: the background model representation, the initialization, and the adaptation. This paper proposes a novel initialization algorithm, able to bootstrap an integrated pixeland region-based background modelling algorithm. The input is an uncontrolled video sequence in which moving objects are present, the output is a pixel- and region-level statistical background model describing the static information of a scene. At the pixel level, multiple hypotheses of the background values are generated by modelling the intensity of each pixel with a Hidden Markov Model (HMM), also capturing the sequentiality of the different color (or gray-level) intensities. At the region level, the resulting HMMs are clustered with a novel similarity measure, able to remove moving objects from a sequence, and obtaining a segmented image of the observed scene, in which each region is characterized by a similar spatio-temporal evolution. Experimental trials on synthetic and real sequences have shown the effectiveness of the proposed approach.