Image Matching Based on Harris-Affine Detectors and Translation Parameter Estimation by Phase Correlation

Abstract

Image matching is an important technology in the filed of computer vision. An effective image matching method based on the Harris-Affine detector and translation parameter estimation by phase correlation is proposed and studied in depth. Firstly, translation parameters of two images are estimated by phase correlation method, and overlapping areas are determined in these two images. Secondly, feature points are detected by using the Harris-Laplace detector in overlapping areas of these two images. Thirdly, these feature points are taken as initial points, and the Harris-Affine detector is used to detect those feature points with affine invariance. Finally, image matching is performed by using a normalized cross-correlation algorithm. Simulation experiments are carried out by using the proposed image matching method. Experimental results demonstrate that the proposed method can match feature point pairs accurately and effectively. The proposed image matching method can be used in the fields of three-dimensional reconstruction, augmented reality and teleoperation robot.