Summary :

Since recent decade, the development of high quality acquisition devices for medical applications yielded migration of traditional video analysis techniques into medical domain. One of the fondamental probelms in Video analysis is motion estimation. Amongst the approaches which have been developped for this purpose, the robust motion estimators are of special interest. Initiated in the works by Black and Anondan, they are now more and more penetrating in the medical imaging and analysis. In the first part of the talk we will present a robust motion estimator from primary optical flow. Developed on the basis of robust estimator of Tukey, it allows for compensation of global motion in a video sequence.  The coherence of motion parameters allow for segmenting of video sequences for the primary  video summarisation. The results will be demonstrated on endoscopic videos. In the second part of the seminar we will be interested in primarily optical flow estimation on MRI sequences of beating heart. Here the motion compensation will be presented for MRI – guided thermotherapy, where precise motion compensation is needed for correction of temperature maps. Finally, we will give some insights in modelling of visual attention for wearable video acquisition set-ups in the studies of Instrulmental Activities of Daily living  of people with dementia for egocentric scenes interpretation.

CV :

Jenny Benois-Pineau is a distinguished professor of Computer science at the University Bordeaux, France and chair of Video Analysis and Indexing research group in Image and Sound Department of LABRI UMR 58000. She is also a deputy scientific director of theme B of French national research unity GDR CNRS ISIS. Her topics of interest include image and video analysis and indexing, motion analysis, content description and machine learning. She is the author and co-author of more than 120 papers in international journals, conference proceedings, book chapters. She has tutored an co-tutored 20 PhD students and 19 research masters. She is associated editor of EURASIP Signal Processing:Image Communication, Elsevier, Multimedia Tools and applications, Springer, She has served in numerous program committees in international conferences and workshops: ACM MM, CIVR, CBMI, AMR, IPTA, SAMT, ECMCS…
She has served as expert for European Commission since FP4 and she is a member of Multimedia Commission of French Ministry of National Education. She has been coordinator or leading researcher in international research projects Platon, Balaton, IP XMedia, Dem@care, French representative in COST292 European action, national research projects ANR and numerous projects with French industrial companies. She gave invited lectures at the universities of Sussex (GB), UPC, UAM (Spain), UNAM (Mexico), UNC at Chapel Hill, CMU, NJIT, Brooklynn Polytechnic (USA), Firenze (Italy), Vesprem (Hungary).

Abstract:

Undoubtedly, the video data is becoming more and more important in a lot of engineering fields, and also in the daily life of human beings. In this talk, two approaches in two different perspectives help achieve the better and deeper understanding of the video data, are introduced, namely, the video key frame selection and its visualization, and the trajectory clustering and visualization. Firstly, A new technique for key frame extraction and 3D visualization, for the sake of video browsing and exploration, is presented. Our proposed approach for key frame selection makes use of the information theory metrics and the f-divergences for calculating the frame-by-frame distance to segment the video clip and then to obtain the key frames. As for the visualization, a 3D interactive interface to display the key frames and their related comprehensive information to provide an expressive exploration of key frames as well as their video contexts is presented. A visualization utility, killTool, used for interactively removing the possible redundant key frames is also developed.
Secondly, a new method for the trajectory clustering and visualization is demonstrated, as an effective way to help observe and understand the (video) motion trajectory data. The main contributions by the proposed method involve the hierarchical clustering in consideration of both speed and direction properties of a trajectory, the introduction of the uncertainty for the updating trajectories based on their speed and direction attributes, and the graphical visualization on the clustered trajectories. In the end, some future considerations are highlighted and discussed.

Short bio:

Qing Xu is a full professor at the School of Computer Science and Technology at Tianjin University, China, where he is also the Director of the Information Technology Research and Civil Aviation Applications Laboratory. He received the M.S. and Ph.D. degrees, respectively in 1998 and 2001, in computer science and technology from Tianjin University, China. His skills include diverse aspects of the information technology, from theory to practice, thanks to acting different roles in different working situations. His research activities cover interdisciplinary fields related to visualization, visual analytics and computer graphics, digital image/video processing, information theory and fuzzy techniques. He has been leading national, international cooperation and industrial research projects, and teaching courses in the technical areas mentioned above. He also serves for the program committee and/or reviewer of a number of journals and conferences in computer graphics, virtual realities, and image/video processing.
Currently, his research interests are in Visual Analytics on Video Data, Analysis of (Video) Motion Trajectories, Monte Carlo based Realistic Image Synthesis, and information theoretic based computational aesthetics.

Abstract:
Weshalb klingen Oktaven völlig rein und können Quinten beim Stimmen von Streichinstrumenten absolut richtig austariert werden? Weshalb werden kleine Sekunden oder das Drei–Ganzton–Intervall (Tritonus) in der Musik als Dissonanzen wahrgenommen? Sind das nur subjektive, beim Hören abendländischer Musik erlernte Empfindungen oder stecken dahinter allgemeine Prinzipien? In diesem Beitrag soll verdeutlicht werden, dass anhand der natürlichen Obertonreihe mathematisch begründet ist, weshalb wir Intervalle musikalisch als konsonant oder dissonant empfinden, weshalb Dur– und Moll–Tonleiter gerade so sind, wie sie heute gebräuchlich sind und weshalb im Laufe der Musikgeschichte die Oktave in 12 Halbtonschritte eingeteilt wurde. Die Obertonreihe ist jedoch mit unserer Empfindung der Tonhöhe als Logarithmus der Schwingungsfrequenz nicht in Einklang zu bringen, was dazu führt, dass ein widerspruchsfreies musikalisches Stimmungssystem nicht existieren kann. Einige historische Kompromissvorschläge zur Stimmung von (Tasten–)Instrumenten werden vorgestellt und verglichen.

CV:
Johannes Huber studierte Elektrotechnik an der Technischen Universität München und erwarb 1977 das Diplom. Er wurde 1982 zur Dr.-Ing. promoviert und erhielt 1991 den Titel Dr.-Ing. habil. mit einer Monographie zur Trelliscodierung. Seit 1991 ist er Professor für Nachrichtentechnik und Leiter des Lehrstuhls für Informationsübertragung an der Friedrich-Alexander-Universität Erlangen-Nürnberg. Von 2007 bis 2009 leitete er deren Technische Fakultät als Dekan.
In der Forschung ist Johannes Huber auf den Gebieten digitale Übertragung, Informations- und Codierungstheorie, codierte Modulation, Entzerrungs- und Detektionsverfahren, MIMO- Übertragungsverfahren etc. aktiv. Er hat zwei Monographien verfasst und ist Autor und Co-Autor von ca. 320 wissenschaftlichen Beiträgen.
Publikationen, die er verfasst bzw. mit verfasst hat, wurden mit dem Preis der deutschen informationstechnischen Gesellschaft in den Jahren 1988, 2000 und 2006 ausgezeichnet. 2004 erhielt er den Innovationspreis der Vodafone-Stiftung für Mobilfunk. In den Jahren 2003 und 2010 wurde ihm der EEEfCOM Innovationspreis verliehen.
Bislang wurden unter der Betreuung von Prof. Huber 41 Promotionen abgeschlossen, 10 seiner ehemaligen Doktoranden sind inzwischen selbst Professoren an namhaften Universitäten und Hochschulen.
Prof. Huber ist Fellow of the IEEE, Corresponding Fellow of the Royal Society of Edinburgh und ordentliches Mitglied der Bayerischen Akademie der Wissenschaften.