Abstract: Default image browsing interfaces on touch-based mobile devices provide limited support for image search tasks. To facilitate fast and convenient searches we propose an alternative interface that takes advantage of 3D graphics and arranges images on a rotatable globe according to color similarity. In a user study we compare the new design to the iPad's image browser. Results collected from 24 participants show that for color-sorted image collections the globe can reduce search time by 23% without causing more errors and that it is perceived as being fun to use and preferred over the standard browsing interface by 70% of the participants.

Abstract: This paper introduces a scalable approach for efficient, low-cost multi-level Quality of Service (QoS) estimation in large overlay networks (ON). We modify an existing distributed partitioning algorithm [1], and use it to create ”QoS maps”. QoS maps empower applications to quickly predict several QoS metrics for any given route, and to obtain multiple alternative routes to any target node in the ON. We show that our modifications of the partitioning algorithm permit the aggregation of large hubs, but still preserve the sublinear runtime of the original heuristic. Simulations with large ONs are performed to evaluate the proposed approach and demonstrate its scalability. Finally, we outline our estimation algorithm that we use to predict QoS and perform QoS aware routing in any given ON.

Abstract: Searching and retrieving videos in a meaningful way on the web is still an open problem. The integration of a user's context into search is one of the most promising approaches to enhance current search interfaces and algorithms. We performed two user studies investigating video search, video retrieval, and video sharing behavior. In this paper we present an overview on the findings most promising for enhancing user experience in video retrieval. We propose a software prototype that implements an adaptive video retrieval system, that utilizes the users' intentions to provide better search results in a user interface adapted to the intentions and needs of users.

Abstract: This thesis discusses three major issues that arise in the context of non-sequential usage of multimedia content, i.e. a usage, where users only access content that is interesting for them. These issues are (1) semantically meaningful segmentation of videos, (2) composition of new video streams with content from different sources and (3) non-sequential presentation of multimedia content. A semantically meaningful segmentation of videos can be achieved by partitioning a video into scenes. This thesis gives a comprehensive survey of scene segmentation approaches, which were published in the last decade. The presented approaches are categorized based on the underlying mechanisms used for the segmentation. The characteristics that are common for each category as well as the strengths and weaknesses of the presented algorithms are stated. Additionally, an own scene segmentation approach for sports videos with special properties is introduced. Scenes are extracted based on recurring patterns in the motion information of a video stream. Furthermore, different approaches in the context of real-life events are presented for the composition of new video streams based on content from multiple sources. Community-contributed photos and videos are used to generate video summaries of social events. The evaluation shows that by using content provided by a crowd of people a new and richer view of an event can be created. This thesis introduces a new concept for this emerging view, which is called ``The Vision of Crowds''. The presentation of such newly, composed video streams is described with a simple but powerful formalism. It provides a great flexibility in defining the temporal and spatial arrangement of content. Additionally, a video browsing application for the hierarchical, non-sequential exploration of video content is introduced. It is able to interpret the formal description of compositions and can be adapted for different purposes with plug-ins.

Abstract: In this position paper we introduce the idea of generating a superior view of a large social event, based on user-generated -- crowdsourced -- content. Instead of just collecting and making them available in a raw form (as social platforms like YouTube), we automatically generate semantically coherent summarizations of the entire event. The individual consuming user gets thus a compact view generated by a large number of producing users. We call this idea the "Vision of Crowds". A case study has been conducted at a social event where we used user-generated content to automatically generate live reports about that event. Furthermore, we have implemented a GUI that allows users to interactively compose personalized video summaries, based on the user-generated data collected at the case study.

Abstract: The way by which multimedia contents are produced, delivered across networks, and consumed by intended users have shifted significantly during the past 10 years. In this paper we postulate that, in the near future, flexible and self-organizing facilities will play a dominating role in distributed multimedia systems. We discuss how such systems can be designed, using a three-layer (sensor, distribution, and user layer) architecture, SOMA (Self Organizing Multimedia Architecture), as an example. We also present innovative directions in three main aspects of self-organized multimedia systems: (i) the self-organizing aspects of multimedia user communities, e.g., the wisdom, intentions, and needs of users; (ii) a fresh look at video streams that treat them as a collection of units that can be composed taking user and network aspects into account; and (iii) new delivery paradigms and how self-organization and multimedia delivery can be combined.

Abstract: Self-organizing overlay networks have received a lot of attention in the recent years. However, despite the popularity of content-aware and topology-aware overlay networks, surprisingly little research has been done to combine both approaches. In this paper, we create robust and flexible overlay networks that we call “Flocks”, which can be content-aware, topology-aware, or a combination of both. We model affinity with interests and properties and show the resulting overlay networks work in a decentralized, self-organizing way, and stabilize quickly.

Abstract: The contributions of this paper are threefold: (1) the extensive introduction of a formal Video Notation (ViNo) that allows for describing different multimedia transport techniques for specifying required QoS; (2) the application of this formal notation to analyzing different transport mechanisms without the need of detailed simulations; (3) further application of ViNo to caching techniques, leading to the introduction of two cache admission policies and one replacement policy supporting nonsequential multimedia access. The applicability of ViNo is shown by example and by analysis of an existing CDN simulation. We find that a pure LRU replacement yields significantly lower hit rates than our suggested popularity-based replacement. The evaluation of caches was done by simulation and by usage of ViNo.

Abstract: In this paper, a novel low level feature for content based image retrieval is presented. The proposed feature structure combines color and spatial color distribution information. The combination of these two features in one vector classifies the proposed descriptor to the family of Composite Descriptors. In order to extract the color information, a fuzzy system is being used, which is mapping the number of colors that are included in the image into a custom palette of 8 colors. The way by which the vector of the proposed descriptor is being formed, describes the color spatial information contained in images. To be applicable in the design of large image databases, the proposed descriptor is compact, requiring only 48 bytes per image. Experiments presented in this paper demonstrate the effectiveness of the proposed technique especially for Hand-Drawn Sketches.

Abstract: We introduce an easy-to-use video browsing tool which assists users in getting a quick overview of videos as well as in finding segments of interest. It provides a parallel and a tree-based view for browsing the content of videos -- or even video collections -- in a hierarchical, non-sequential manner. The tool has a plug-in architecture and can be extended both by further presentation methods and by video analysis algorithms.

Abstract: We present an algorithm for video scene detection based on the identification of recurring motion sequences within a video stream. The motion information is extracted in the compressed domain of H.264/AVC videos, no full decoding of the video stream is needed. Based on the motion information our algorithm identifies sequences of adjacent frames with similar motion. Throughout all identified motion sequences we are searching for recurring patterns of similar ones. The most recurring pattern is used for the segmentation of the video stream into scenes. The evaluation shows promising results.

Abstract: We present a new approach for video browsing using visualization of motion direction and motion intensity statistics by color and brightness variations. Statistics are collected from motion vectors of H.264/AVC encoded video streams, so full video decoding is not required. By interpreting visualized motion patterns of video segments, users are able to quickly identify scenes similar to a prototype scene or identify potential scenes of interest. We give some examples of motion patterns with different semantic value, including camera zooms, hill jumps of ski-jumpers, and the repeated appearance of a news speaker. In a user study we show that certain scenes of interest can be found significantly faster using our video browsing tool than using a video player with VCR-like controls.

Abstract: We present a video browsing tool that uses a novel and powerful visualization technique of video motion. The tool provides an interactive navigation index that allows users to quickly and easily recognize content semantics like scenes with fast/slow motion (in general or according to a specific direction), scenes showing still/moving objects in front of a still/moving background, camera pans, or camera zooms. Moreover, the visualization facilitates identification of similar segments in a video. A first user study has shown encouraging results.

Abstract: A new approach for interactive video browsing is described. The novelty of the proposed approach is the flexible concept of interactive navigation summaries. Similar to time sliders, commonly used with standard soft video players, navigation summaries allow random access to a video. In addition, they also provide abstract visualizations of the content at a user-defined level of detail and, thus, quickly communicate content characteristics to the user. Navigation summaries can provide visual information about both low-level features but even high-level features. The concept fully integrates the user, who knows best which navigation summary at which level of detail could be most beneficial for his/her current video browsing task, and provide him/her a flexible set of navigation means. A first user study has shown that our approach can significantly outperform standard soft video players - the state-of-the art “poor man’s” video browsing tool.

Abstract: Commercial blocks provide no extra value for video indexing, retrieval, archiving, or summarization of TV broadcasts. Therefore, automatic detection of commercial blocks is an important topic in the domain of multimedia information systems. We present a commercial detection approach which is based on logo detection performed in the compressed domain. The novelty of our approach is that by taking advantage of advanced features of the H.264/AVC coding, it is both significantly faster and more exact than existing approaches working directly on compressed data. Our approach enables removal of commercials in a fraction of real-time while achieving an average recall of 97.33% with an average precision of 99.31%. Moreover, due to its run-time performance, our approach can also be employed on low performance devices, for instance DVB recorders.

Abstract: MPEG-7 is an extensive multimedia metadata standard covering a huge number of aspects of metadata. However, as with most metadata standards details of usage and application of the standards are – at least partially – open to interpretation. In case of MPEG-7storage and transmission of high level metadata on concept level are defined but retrieval methods are not proposed. So if for instance a user annotates photos using the MPEG-7 semantic description scheme, there are no standardized ways to retrieve the photos based on the annotation. In this paper we propose metrics for retrieval based on the MPEG-7 semantic description scheme and evaluate them in a digital photo retrieval scenario.

Abstract: Arthroscopic surgery is a minimally invasive procedure that uses a small camera to generate video streams, which are recorded and subsequently archived. In this paper we present a video summarization tool and demonstrate how it can be successfully used in the domain of arthroscopic videos. The proposed tool generates a keyframe-based summary, which clusters visually similar frames based on user-selected visual features and appropriate dissimilarity metrics. We discuss how this tool can be used for arthroscopic videos, taking advantage of several domain-specific aspects, without losing its ability to work on general-purpose videos. Experimental results confirm the feasibility of the proposed approach and encourage extending it to other application domains.

Abstract: The increasing availability of short, unstructured video clips on the Web has generated an unprecedented need to organize, index, annotate and retrieve video contents to make them useful to potential viewers. This paper presents a novel, simple, and easy-to-use tool to benchmark different low level features for video summarization based on keyframe extraction. Moreover, it shows the usefulness of the benchmarking tool by developing hypothesis for a chosen domain through an exploratory study. It discusses the results of exploratory studies involving users and their judgment of what makes the summary generated by the tool a good one.