A new research of sub-pixel level accuracy of TDICCD remote sensing image registration

In the field of remote sensing imaging, TDICCD remote sensing images have a lot of their own characteristics, such as high-resolution, large amount of information, less overlapping parts of pixels, additional image blurring etc. Therefore, there exist many difficulties, especially in terms of high-accuracy registration of pairs of images. For that, this paper presents two new pixel interpolation method for sub-pixel level registering images that allows for scaling, translation and rotation. The proposed technique, which is based on the maximization of the correlation coefficient function, combines an efficient pixel-moving interpolation scheme with surface fitting, which greatly reduces the overall computational cost. The accuracy of the algorithm is evaluated by calculating correlation coefficient of couples of points belonging to images transformed with preset factors and also comparing it to other sorts of methods. The experiment results show that the accuracy of registration reaches 0.01 pixels.