High-Resolution Remote-Sensing Image Registration Based on Angle Matching of Edge Point Features
Qing Guo, Mengmeng He, An Li · IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing · 2018
Remote-sensing image registration is an important prerequisite for many remote-sensing applications. The accuracy, efficiency, and automatic degree of registration have direct impacts on follow-up applications. With the improving of remote-sensing resolution, the size and the data amount of remote-sensing image are constantly increasing. Meanwhile, with the development of remote-sensing application, the performance of registration is also put forward increasing requirement. Therefore, this paper proposes an automatic and fast image registration method based on the angle matching of edge point features (EPFs). First, the original image is transformed by the Haar wavelet to get the approximate image to improve the registration speed. Second, edges in different source images are extracted by different edge detectors. Then, edge line features are transformed into EPFs to get accurate positioning and stable features. Third, the initial matching point pairs (MPPs) are determined by the ratio-the minimum angle to the second minimum angle between feature vectors of EPFs-which is less than a threshold. Fourth, in order to enhance the registration accuracy, the random sample consensus is improved by adding the quality constraint condition to delete error MPPs. Finally, the block thought is used to uniformly choose MPPs to calculate the affine transform parameters, which avoids the local optimal problem and further improves the registration accuracy. Multiple high-resolution remote-sensing datasets of registration experimental results indicate that the proposed method is fast implemented and has high accuracy and effectiveness in processing efficiency.