Sub-pixel Image Registration Based on Physical Forces Ali Ghayoor Dept. of Electrical Eng., M.Sc. Student Iran University of Science and Technology Tehran, Iran aghayoor@ee.iust.ac.ir Saeed Ghorbani Dept. of Electrical Eng., M.Sc. Student Sharif University of Technology Tehran, Iran saeid_ghorbani@ee.sharif.ir Ali Asghar Beheshti Shirazi Dept. of Electrical Eng., Faculty Member Iran University of Science and Technology Tehran, Iran abeheshti@iust.ac.ir Abstract—A new method for image registration has been previously proposed by the authors, which the registration is based on physical forces. The registration parameters are translation and rotation. This method assumes images like charged materials that attract each other. In this case, one of the images moves in the same direction as the applied force while the other one is still. The movement of the image continues until the resultant force becomes zero. This approach estimates the registration parameters simultaneously and leading to a better optimized set of registration parameters. The registration error for this method is 1 to 3 pixels. In this paper we aim to develop this method for the applications which need sub-pixel accuracy. First, by applying the Canny edge detector on the input images, the edge information is also used for the registration process to increase the robustness of this method in the presence of noise. After that, sub-pixel accuracy is provided for this method by using interpolation techniques. Keywords- Image registration, Pattern recognition, Area- based methods 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. Thus, the registration techniques can be implemented to integrate complementary information acquired by different imaging modalities. It geometrically aligns two images—the reference and sensed images. Image registration is widely used in remote sensing, medical imaging, computer vision, etc. Zitova et al. [1] surveyed the recent image registration techniques covering different application areas. During the past two decades, many registration methods have been developed, which can be categorized by the spatial transformation (rigid or non-rigid registration), dimension (2D or 3D), similarity criterion (sum of square differences(SSD), cross correlation(CC), or mutual information(MI)), and optimization method (exhaustive search, Powell`s method, and Simplex). However, the proposed method [2] is not included in any of these categories, as it implies the similarity criterion and the optimization method simultaneously, which is a unique feature of this method. However, if we divide the feature matching methods into two categories: the area-based methods and the feature-based methods, the proposed method [2] falls into the first category. In despite of feature-based methods, area-based methods deal with the images without attempting to detect salient objects. Area-based methods, sometimes called correlation- like methods or template matching, [3] merge the feature detection step with the matching part. Classical area-based methods are cross-correlation (CC) methods, Fourier methods and the mutual information (MI) methods. Cross-correlation [4] exploits the image intensities directly, without any structural analysis. This measure of similarity is computed for window pairs from the sensed and reference images and the registration parameters are searched to maximize it. The window pairs for which the maximum is achieved are set as the corresponding ones. Although the CC based registration can exactly align translated images only, it can also be successfully applied when slight rotation and scaling are present. There are generalized versions of CC for more geometrically deformed images. Two main drawbacks of the correlation-like methods are the flatness of the similarity measure maxima (due to the self-similarity of the images) and high computational complexity. However, despite these limitations, the correlation-like registration methods are still often in use, particularly thanks to their easily hardware implementation, which makes them useful for real-time applications. Various registration techniques based on Fourier Transform have been proposed. They are based on the following three properties: shifting (translation), rotation and scaling properties [5]. They exploit the Fourier representation of the images in the frequency domain. The phase correlation method is based on the Fourier Shift Theorem and was originally proposed for the registration of 978-1-4244-7555-1/10/$26.00 ©2010 IEEE