Infrared Small Target Detection via Global Motion Compensation and Concave Curvature–Based Motion Modeling

Infrared small target detection under complex backgrounds remains a challenging task due to the high similarity between background clutter and true targets, which often leads to failure in clutter suppression and target detection. This paper proposes a novel infrared small target detection method under complex backgrounds and moving camera conditions. The method comprises two synergistic components— a spatial enhancement algorithm named Multi-scale Local Isotropic Concave Normal Curvature (MLICNC) and a temporal association strategy called Global Motion Compensation based Motion Outlier Detection (GMC-MOD). MLICNC enhances target saliency by leveraging the isotropic concave normal curvature property and suppresses surrounding background clutter via a multi-scale local contrast mechanism. In the temporal domain, GMC-MOD performs frame registration using homography transformation and conducts local window-based matching in conjunction with Kalman filtering for trajectory association and motion analysis. By identifying motion-consistent outliers within a globally compensated coordinate system, the method achieves robust and accurate detection performance. Experimental results on real-world infrared sequences demonstrate that the proposed method significantly improves detection capability compared with existing approaches.