Research

The laboratory carries out research in computer vision, multimedia processing, management and security, as well as human-machine interaction. Research applies to visual media, such as images and videos, to sounds and to biological signals. Research results are described in over 300 refereed scientific publications and in 8 patents. Roughly 75% of the research is financed by external, competitive grants.

Content-based Visual Indexing and Retrieval. Image and video archives: current multimedia systems require efficient algorithms for archival and retrieval of images and image sequences, for both individual and professional users. Database strategies are being developed for efficient retrieval of documents on the basis of textual and visual criteria. Several applications have been developed or are currently underway (museum archives, Distributed Video Production, Viper )

Stochastic Image Processing. Image and video watermarking: copyright protection mechanisms for images and videos are being developed, by means of spread-spectrum-based digital watermarks embedded in the documents and a secure copyright network allowing the management of the copyright certificates. We are also currently researching benchmarking of algorithms within the Checkmark project.

Multimodal Interaction. The Multimodal Interaction group (MMI) aims at studying several forms of interaction of human with computers and machines, in addition to the classical screen-based visual mode. The interactions modalities currently considered are: tactile, auditory and recently biological signals.

On the applied side, three main projects have now been completed, leading, respectively, to

• AB-Web: a prototype WWW browser for visually impaired and blind users is under development, including text-to-speech and 2D image-to-3D sound conversion.
• a public-domain software for image processing (LaboImage)
• a machine vision system for agricultural robotics (Potato Operation)
• a medical classification system for 2D gel images (as part of the Melanie system)

On the theoretical side, at various times the following topics have been investigated:

• image filtering and segmentation
• low-level grouping
• geometric invariance
• motion analysis
• visual attention
• object recognition and learning
• computational neuroscience

Most of CVML research is financed by external, competitive grants:

• Swiss: National Competence Center in Research “IM2 – Interactive Multimodal Information Management”, National Research Foundation, Hasler Foundation, Commission pour la Technologie et l’Innovation, Virtual Campus, Priority Research Program for Information and Communication Structures, Priority Research Program in Computer Science, National Research Program “AI and robotics”;

• European: ”Peer-to-Peer Tagged Media (PetaMedia) Network of Excellence”, “Multimatch”, “Similar - Network in human-machine interfaces and communication”, “Ecrypt - Network in cryptology and watermarking”, “Webkit – Intuitive physical interfaces to the WWW”, “M4 – Multimodal Meeting Manager”, “Certimark – Certification for watermarking techniques”, “JEDI-FIRE – JAVA Extensive Software Defined Internetworking – Flexible Electronic Commerce Firewall and Data Privacy”, “DVP – Distributed Video Production”;

• Private companies and public associations (Association pour le Bien des Aveugles et Amblyopes).

The following research grants have been obtained from the European Union, from the Swiss National Research Foundation (FNRS), and from various Swiss National Research Programs (the Group has also been associate partner in other grants from the FNRS).