University of Konstanz
Graduiertenkolleg / PhD Program
Computer and Information Science

Colloquium of the Department and the PhD Program


Stochastic methods for finding image structures


Prof. Dr. Tamás Szirányi, Computer and Automation Inst., Hungarian Academy of Sciences
Budapest, Hungaria

date & place

Wednesday, 16.12.2009, 16:15 h
Room C 252


Retrieving geometrical information from images or videos by statistical evaluation of elementary image information.

We search for image structures of videos by using stochastic methods without any preliminary object definition: correlating changes in different views and estimation of relative focus map.

First, a motion-based method is presented for automatic registration of images in multi- camera systems, to permit synthesis of wide-baseline views. We can register image views by simple motions like quivering of the leaves on a tree... Another interesting application is: where is the mirror in an image; we find the vanishing-point position for planar reflected images or shadows.

An automatic solution for finding focused areas based on localized blind deconvolution is also presented. This approach makes possible to determine the relative blur of image areas without a priori knowledge about the images or the shooting conditions. The method uses iterative feature evaluation and a new type of error measure for classification. We also show the methodsī usability for applications in content based indexing and video feature extraction.