A coarse-to-fine heterologous registration method for Infrared-Visible images based on MDC and MSMA-SCW descriptors

Heterologous image registration is one of the key technologies for information fusion of infrared image and visible image. Considering the differences of image resolution, feature, and viewing angle between infrared image and visible image, a method consisting of MDC (Multi-scale Duty Cycle) descriptor and MSMA-SCW (Multi-Scale and Multi-Angle partitioned for Shape Context Weighted) descriptor based on edge feature description is proposed for heterologous image registration. The MDC descriptor is designed to describe the edge feature of image, which is used for coarse registration to find the maximum similarity region and primary matching scale. The MSMA-SCW descriptor was designed to describe the key point on the maximum similarity region, which is used for fine registration to match the key points at refined multi-scale between two images. Finally, the transformation relationship of image position between the two images could be acquired through affine transformation based on the results the matched key points. Experimental results have shown that the proposed method is of high accuracy in heterogeneous images registration of infrared images and visible images.