| |||||||||||
|
Acivs 2008 Tutorial on Similarity measures for satellite image registration
Acivs 2008 features a pre-conference tutorial on "Similarity measures for satellite image registration". The tutorial will take place on Monday October 20, 2008, between 14:00 and 17:00. To attend this tutorial you need to check the appropriate box on the registration form. The price of the tutorial is 80 euro.
Lecturer
Jordi Inglada. CNES DCT/SI/AP. Toulouse. France
Abstract
The problem of image registration is usually addressed as optimization of a deformation between images optimizing a given similarity measure.
Similarity measures as Mutual Information are becoming a classical choice for the problem of comparing images acquired in a multi-sensor context. They are a good choice when the link between the two sets of data to be compared is not known or can not be simply modeled. However, since these measures are based on the estimation of probability densities, they are difficult to implement when the size of the expected changes is small. Also, when the shape of the changes is not suited to a classical rectangular estimation window, the computation of the similarity may be difficult.
With the arrival of new high and very high resolution imaging sensors, the heterogeneity inside the information classes and the ability to distinguish small objects of interest will have to be accounted for. In this context salient points which can be linked to invariants in the landscape, can be used for image registration.
In this tutorial, after an introduction to classical multi-modal similarity measures, several new challenges related to the processing of high and very high resolution image data will be addressed and some insights of interesting research problems will be presented.Theory and applications will be illustrated by means of interactive demonstrations.