Machine efficient adaptive image matching based on the nonparametric transformations

Abstract

The paper presents machine efficient area-based image matching method that is based on a concept of matching-regions that are adaptively adjusted to image contents. They are in a form of square windows which grow to convey enough information for reliable matching. This process is controlled by local image contents. The images, however, are transformed into nonparametric representation. Such a liaison of information-theoretic models with nonparametric statistics allows for compensation for noise and illumination differences in the compared images, as well as for better discrimination of compared regions. This leads to more reliable matching in effect. Machine efficient implementation is also discussed in the paper. Finally the experimental results and conclusions are presented. © Springer-Verlag Berlin Heidelberg 2005.