A Novel Image Registration Algorithm in Aerial Infrared Videos

This paper focuses on image registration between sequential frames in aerial infrared video. To address this problem, we propose an efficient approach based on an improved Smallest Univalue Segment Assimilating Nucleus (SUSAN) corner detector and a novel corner rejection strategy. Firstly, the SUSAN corners are detected via the concatenation of enlarging the radius of original detector and down-sampling. Then, the corner rejection strategy is applied to remove the corners on moving objects. After that, Histograms of Oriented Gradients (HOG) is used to form the descriptor of each corner, followed by a feature matching phase to obtain correspondences of corners. Finally, RANdom Sample Consensus (RANSAC) is performed to estimate the parameters of the transformation model with the correspondences. Experimental results show that our approach is fast and stable for image registration in aerial infrared videos.