Robust sub-pixel image registration based on combination of local phase correlation and feature analysis

This paper presents a robust sub-pixel image registration algorithm by analyzing intensity variation of local features. The proposed method consists of three steps: i) global image registration using features analysis, ii) local image registration using phase correlation, and iii) sub-pixel image translation. It can register two images without mis-registration artifacts because of the accurate estimation of inter-image transformation estimation by using local phase correlation and feature analysis. The proposed algorithm can be applied for various image processing systems including image registration, video stabilization, and multi-sensor image fusion.