Infrared Small Target Detection via Spatial-Temporal Total Variation Regularization and Weighted Tensor Nuclear Norm

The infrared small and dim targets are often buried in strong clutters and noise, which requires robust and efficient detection approaches to achieve search and track task. In this paper, a novel infrared small target detection approach based on tensor robust principal component analysis (RPCA) is proposed by fully utilizing both spatial and temporal information. Traditional total variation (TV) regularization-based method only considers the spatial information of a single frame, and its efficiency cannot meet the real-time processing requirement. We incorporate an anisotropic spatial-temporal TV regularization to utilize both single-frame and inter-frame information. First, the input image sequence is transformed into an infrared patch tensor (IPT) model. Then, the spatial-temporal TV regularization and weighted IPT model (STTVWNIPT) is proposed to separate the target and background. For the low-rank background component, we adopt the weighted tensor nuclear norm and spatial-temporal TV regularization to describe the smoothness. For sparse target component and a noise component, we use the $l_{1}$ norm and Frobenius norm terms to characterize, respectively. Finally, the proposed model can be solved efficiently by the alternating direction method of multipliers (ADMM). The extensive experiments demonstrate that the proposed model outperforms the other competitive methods, and it is much more efficient than traditional TV-based method.