Saturday, July 19, 2008

France PostDoc Position on Saliency Maps

France: PostDoc Position on Saliency Maps for Video (Saint Etienne, France and Gjovik, Norway)

Please kindly mention ScholarshipNet when applying for this position
PostDoc position on saliency maps for video (Saint Etienne, France and Gjovik, Norway)

Title: Extension of the saliency maps to video
Under natural viewing conditions (no tasks assigned to viewers) humans tend to fixate on specific parts of an image that are of interest to them naturally. These regions carry most of the useful information needed for our interpretation of the scenes we encounter in our daily life activities. Therefore, numerous applications in the field of engineering, marketing and art can benefit from understanding the human visual attention mechanisms, such as visual media quality evaluation, advertisement design, graphical user interface design, etc. .

Several models are proposed in the literature for finding visually important regions in still images, based on the notion of saliency. Basically, a saliency map is the map of regions which are more prominent than others in terms of low level image properties such as intensity, color and orientation. This could be a very crude and simplistic way of approximating the human visual attention mechanisms; since it is known that familiar objects and especially faces present in an image are strong factors in drawing our attention to certain regions of an image.

This was the motivation for NCRL’ work on including face detection as an integral part of the new attention model Mr. Puneet Sharma has been working on for his MSc thesis at NCRL. His work on the automatic computation of saliency maps using low-level features such as intensity, color and orientation combined with high-level features was very promising and generated couple of publications. The objective of this research was to understand the factors that influence the saliency and gaze maps and to modify the saliency map extraction algorithm in order to make the obtained saliency maps more similar to the experimental gaze maps. Puneet has shown that we can achieve around 30% on average for the set of images used in the correlation between the automatically computed saliency maps and the experimentally generated gaze maps based on the new model.

An obvious continuation of this work is to extend this model to analyze video content. A straight forward way of doing that is by including a motion analysis module in the modified model. Dr. Marco Carli worked on a related problem, where he analyzed video content for the automatic evaluation of the perceived video quality. In his work he did not investigate in depth the use of low-level features such as color and did not use high-level objects detection as we did for the faces. He did however focus on motion information analysis and their contribution to the saliency of certain objects or regions in video. Therefore, a combination of both NCRL’ model with the face detection module and Dr. Carlis’ motion information processing would create a better modeling of the HVS attention mechanisms.

This PostDoc will be funded by the University of Saint-Etienne (France).

The candidate selected will work in the Laboratory Hubert Curien at Saint Etienne (France) under the supervision of Professor Alain Trémeau and at Gjovik (Norway) under the supervision of Dr. Faouzi Alaya Cheikh ( ).

We seek a strong and motivated candidate. Applicants should submit a CV including a brief research statement and a list of publications. Selection will be based on: 1. Excellence of the candidate: outstanding achievement in the applicant’s PhD degree level in imaging science, computer science, or any discipline pertaining to the quantitative description of image processing. 2. Language ability: the candidates must demonstrate sound knowledge of the language. 3. Candidate motivation to undertake the PostDoc and relevance to his/her professional development (explaining the application, the present situation, the interest in the PhD degree, the intentions after this PhD degree, …).

The application deadline is September 8, 2008 .

To get more information contact: Prof. Alain TREMEAU, Alain.Tremeau[ at ]