Investigating two major issues in change detection: Misregistratoin and thresholding

Methods of change detection are increasing dramatically in the last few years since timely and accurate change detection of the earth surface is of high importance in order to make better decision. The purpose of this research is to investigate two major issues in change detection: first, the effect of image misregistration and secondly, the effect of applying different thresholding algorithms in classical change detection techniques. Moreover a new method of automatic change detection is introduced. At first, misregistration and its effect on change detection results is investigated comprehensively in a quantitative manner. At a later stage the problem with classical change detection procedures is indentified and an approach based on Markov Random Fields using contextual information is introduced.

Keywords:

Change detection , misregistration , thresholding , Markov Random Fields