Accurate point matching based on combined moment invariants and their new statistical metric

In this paper, an accurate point matching scheme based on Combined Moment Invariants (CMIs) and their new metric is presented. In general, the matching of the local similarity detection by the combined invariants and conventional distances produces some outliers, which should be deleted firstly through some complex statistics. In order to obtain the more reliable matching results, we construct a new metric for combined NMIs. The whole framework involves two steps: 1) Extraction of Control points (CPs) on the reference image _the canny edge detector and well-known Harris detector are described to extract the edges and corner points. 2) Searching for the corresponding CPs in a circular of the matched image_is based on local similarity metric with combined NMIs. The framework is fully automatic and simple without any additional steps. It has been successfully applied to register remote sensing images. Experimental results show that the proposed scheme excludes the outliers successfully for their high matching accuracy.