Infrared Small Target Detection via Spatial-Temporal Infrared Patch-Tensor Model and Weighted Schatten p-norm Minimization

Abstract Infrared small and dim target detection has been a typical application of infrared imaging techniques, and it still remains a challenging problem to achieve robust performance and fast processing capability under complex clutters and heavy noise. To cope with these obstacles, a novel detection method is proposed in this paper. Firstly, we construct a novel spatial-temporal infrared patch-tensor (STIPT) structure for mining valuable information in the time domain, and the small target can be detected after recovering and segmenting the low-rank background tensor. Then, considering the overshrinkage problem in low-rank component estimation field caused by vanilla nuclear norm and weighted nuclear norm, weighted Schatten p-norm is incorporated to improve the performance by considering the particular meaning of different singular values. A solution framework is proposed via Alternating Direction Method of Multipliers (ADMM), then an adaptive threshold is utilized to separate the targets. We take systematic analysis on real infrared data by the qualitative method and quantitative method, and the effectiveness and robustness of this method are verified in various scenarios.