Infrared target detection based on 5-D spatial-temporal knowledge

Infrared (IR) small target detection problem has attracted increasing attention. Tensor theory-based detection techniques have been widely utilized, while facing challenges such as tensor structures, background and target estimation. This paper proposes an IR dim and small target detection method based on 5-D spatial-temporal knowledge (5D-STD). Specifically, a 5-D whitened spatial-temporal patch-tensor is constructed. Then, we design a 5-D tensor nuclear norm for background estimation and a Moreau envelope-derived sparsity estimation norm. Furthermore, we establish a low-rank and sparse decomposition model with an alternating direction method of multipliers (ADMM)-based optimization scheme for IR target detection. Extensive experiments conducted on three real IR sequences prove the superiority of 5D-STD in terms of target detectability, background suppressibility and overall performance.