Global Sparsity-Weighted Local Contrast Measure for Infrared Small Target Detection

Zhaobing Qiu,Yong Ma,Fan,Jun Huang,Lang Wu

DOI: https://doi.org/10.1109/lgrs.2022.3196433

IF: 5.343

2022-01-01

IEEE Geoscience and Remote Sensing Letters

Abstract:Local contrast measure (LCM) proves effective in infrared (IR) small target detection. Existing LCM-based methods focus on mining local features of small targets to improve detection performance. As a result, they struggle to reduce false alarms while maintaining detection rates, especially with high-contrast background interference. To address this issue, this letter proposes global sparsity (GS)-weighted LCM (GSWLCM), which fuses both global and local features of small targets. First, robust LCM (RLCM) is proposed to remove low-contrast backgrounds and extract candidate targets. Then, to suppress high-contrast backgrounds, we customize the random walker (RW) to extract candidate target pixels and construct the global histogram and calculate GS. Finally, GSWLCM fusing global and local features is calculated and the target is detected by adaptive threshold segmentation. Extensive experimental results show that the proposed method is effective in suppressing high-contrast backgrounds and has better detection performance than several state-of-the-art methods.