|Visual Saliency from adaptative whitening|
Seminar by Xosé R. Fdez-Vidal
Abstract: The term visual saliency is widely used in computer vision to refer to any aspect of a stimulus that, for any of many reasons, stands out from the rest. There is a number of applications in which the saliency map is typically used to prioritize selection, e.g. to identify the most important information in visual input streams and to use this to improve performance in late states (scene uderstanding, object recognition, etc.) or transmitting visual data (compression, video resume, thumbnailing, etc.). Our approach to visual saliency that relies on a contextually adapted representation produced through adaptive whitening of color and scale features. Unlike previous models, the proposal is grounded on the specific adaptation of the basis of low level features to the statistical structure of the image. Adaptation is achieved through decorrelation and contrast normalization in several steps in a hierarchical approach, in compliance with coarse features described in biological visual systems. The model is able to predict a wide set of relevant psychophysical observations and predicting human fixations using diferents different eye-tracking datasets.
Short bio: XOSÉ R. FDEZ-VIDAL received the M.S. and Ph.D. degrees, both in Physics, from the University of Santiago de Compostela in 1991 and 1996, respectively. Since 1992, he has been with the Applied Physics Department at Santiago de Compostela University, where he is now an Associate Professor.His primary research interest is in computational models of biological vision, with applications to machine vision. Dr. Fdez-Vidal is a member of IAPR.