Accuracy estimation of images registration method

Accuracy estimation of image registration is important for image fusion, computer vision and so on image processing tasks. It is a substantial part of image registration process. Without quantitative accuracy estimation, no registration algorithm can be applied pragmatically. In this paper, the parameters to measure accuracy of image registration and their characteristics are introduced systematically. And Cross Validation is used to estimate the parameters of measuring accuracy of image registration. And it is used to evaluate accuracy of SIFT algorithm and manual image registration algorithm. The experimental results show the effectiveness of Cross Validation method.

I. Introduction

Image registration is the process of overlaying two or more images of the same scene taken at different times, from different viewpoints and/or by different sensors[1]. Image registration is an important step for computer vision, 3D reconstruction, video surveillance etc. For any registration method and any application field, it is highly desirable to provide the user with an estimate how accurate the registration actually is. Without quantitative evaluation, no registration method can be accepted for practical utilization.