Bilinear Low-rank Matrix Factorization for Infrared Small Target Detection

In recent years, low-rank-based methods have high effectively in infrared small target detection. However, the limitations of low-rank based methods in rank function approximation affect the accurate evaluation background. To more accurately evaluate background, we propose a novel bilinear low-rank matrix factorization method for infrared small target detection. Considering that the bi-nuclear quasi-norm is closer to the rank function, and it can be calculated by the nuclear norms of two smaller factor matrices. In our method, the bi-nuclear quasi-norm is used to constrain low-rank features in infrared images for more accurate background evaluate. To find the solution of the proposed method, we developed an effective algorithm based on alternating direction method of multipliers (ADMM). Extensive experiments demonstrate the superiority of the BLRMF method over mainstream infrared detection methods.