Reweighted infrared patch image model for small target detection based on non-convex ℒp-norm minimisation and TV regularisation

• Published in: IET image processing 2020-07, Vol.14 (9), p.1937-1947
• Main Authors: Rawat, Sur Singh, Verma, Sashi Kant, Kumar, Yatindra
• Format: Article
• Language: English
• Subjects: background patch image; background suppression ability; baseline methods; complex background; image denoising; infrared detection system; infrared images; infrared imaging; infrared patch image model; minimisation; NNM; nonconvex ℒp‐norm minimisation; nuclear norm minimisation; object detection; Research Article; reweighted IPI model; synthetic images; target detection; TV regularisation term
• ISSN: 1751-9659; 1751-9667; 1751-9667
• DOI: 10.1049/iet-ipr.2019.1660

Infrared small target detection in a complex background has always been a challenging task in an infrared detection system. The existing methods based on the infrared patch image (IPI) model have achieved a good result but are sensitive to the complex background. So, to effectively detect the small target in complex background, model based on the reweighted IPI model along with total variance (TV) is proposed in this study. In this study firstly, the problem of using nuclear norm minimisation (NNM) in the existing IPI-based methods is discussed, and a solution is proposed by replacing the existing NNM with the ℒp-norm minimisation of singular values in the existing IPI methods. Secondly, a TV regularisation term is added to the background patch image to suppress the noise and preserve the strong edges in the background. The proposed method is solved by the alternating direction method of the multiplier. The robustness of the proposed method is validated by experimenting with the large dataset of real infrared images as well as the synthetic images. The proposed method not only has good background suppression ability, but also enhances and detect the target well in comparisons with the other baseline methods.