Sub-Pixel Image Registration Based on Extended Alignment Metric

This paper proposes a sub-pixel registration method based on analytic method to meet the requirement of high accuracy and time efficiency in image processing. Image registration is a coarse-to-fine process, thus in this paper, integer pixel match is firstly obtained via maximizing the alignment metric, and the sub-pixel image registration is a single step to solve a optimization problem by analytic method. A mathematical model of estimating image similarity is established with the precise sub-pixel coordinates unknown and a polynomial is constructed to express the similarity between two images. Registration position is the right point the polynomial gets a peak value. Experimental results show the proposed algorithm reduces the computation time and gets a high accuracy when experiments are conducted on image databases and self-collected videos and pictures.