Graph-Regularized Laplace Approximation for Detecting Small Infrared Target Against Complex Backgrounds

Abstract

Against complex background containing the tiny target, high-performance infrared small target detection is always treated as a difficult task. Many low-rank recovery-based methods have shown great potential, but they may suffer from high false or missing alarm when encountering the background with intricate interferences. In this paper, a novel graph-regularized Laplace low-rank approximation detecting model (GRLA) is developed for infrared dim target scenes. Initially, a non-convex Laplace low-rank regularizer instead of the nuclear norm is employed to boost the accuracy of heterogeneous background estimation. Then, to maintain the intrinsic structure between background patch-image, the graph regularization is incorporated in the detecting model. Besides, aiming at reducing the nontarget outliers, a reweighted l 1 norm with nonnegative constraint is used. Finally, the proposed model is extended to a generalized framework (G-GRLA) by replacing different non-convex rank functions. With the help of the alternating direction method of multiplier (ADMM), the solution of the proposed model is obtained by an iterative optimization scheme. The experimental results on extensive actual infrared images present the superior performance of our proposed method to compare with the state-of-the-art methods. INDEX TERMS Small target detection, nonconvex low-rank regularizer, manifold information, graph regularization, reweighted l 1 norm with nonnegative constraint.

A Method for Infrared Sea-Sky Condition Judgment and Search System: Robust Target Detection via PLS and CEDoG

“…Local Contrast Method [17]- [20] Spatio-temporal Saliency Approach [21]- [22] Low-rank Tensor Completion [23]- [24] Based on Deep Learning [25]- [29] It can be known from above scientific research situation that these current methods are more suitable for targets with a single feature in the entire image [30][31][32][33][34]. On the other word, sea-sky background clutter and wave noise in the space or transform domain are irrelevant to the target [35][36][37][38][39][40][41].…”

Robust Infrared Small Target Detection via Jointly Sparse Constraint of l1/2-Metric and Dual-Graph Regularization

“…For the model, one of the limitations is long consuming time due to the slow convergence of the optimization based on accelerated proximal gradient (APG). The alternating direction method of multiplier (ADMM) optimization scheme is used in the models proposed recently since it can get same optimal solution under faster convergence [21,22], [45,46]. To mine more nonlocal self-correlation information from patch models, some extended versions of the IPI model have been proposed, including infrared patch-tensor model (IPT) [21], spatial-temporal patch-image model (STPI) [7], spatial-temporal tensor model (STTM) [47], multi-subspace learning model (SMSL) [20].…”

“…Besides, in the enhancement type of the IPI model, Wang et al [45] used the total variation regularization (TVPCP) to depict the background feature, which aimed to obtain good target-background separation for some mild situations. In [46], we proposed a combination of nonconvex rank approximation and graph regularization (GRLA) to take full use of the intrinsic structure between patch images.…”

Modified Graph Laplacian Model With Local Contrast and Consistency Constraint for Small Target Detection

“…Besides, graph Laplacian has also been employed in some component analysis-based methods. Zhou et al [40] proposed a combination of a reweighted L 1 -norm for the target component and a graph Laplacian regularization for the background component. Later on, they improved their method with a spatial-feature graph Laplacian regularization and a L 1/2 -norm for the target component [41].…”

“…Third, we take advantage of the graph Laplacian's properties and utilize it to enhance the global rarity of small targets. In contrast, researches [40] and [41] take the graph Laplacian as regularization terms to moderate the sensitivity to defective pixels in the component decomposition. Finally, and most importantly, we integrate the local and global priors more expressively and efficiently.…”