A Multi-sensor image registration, method adapted for larger scale

摘要

A new method based on the edge curvature areaextreme-value and the edge correlation for efficient imageregistration is proposed. First of all, draw some edges from thepending image,- set the- edge length threshold by using acomparatively large Gauss parameter,-save the longer edgesonly, and mode them with improved Freeman. Secondly, withthe Freeman code aforementioned, compute the curvature ofthe curve, and match the edges according to the correspondingsituation of the curve area curvature and get the scaletransform parameter. Thirdly, resample the image by usingthat scale parameter, draw some edges again and by applyingthe edge correlation and consistency check means, get thecorresponding pairs of edges. Finally, by using the Least MeanSquare, obtain the registration parameter of the image. Thisarithmetic applies in situations like registration images withcomparatively rotating angles and translational motions.