Infrared small target detection based on variance difference weighted three-layer local contrast measure

Abstract

Infrared (IR) small target detection technology is of paramount importance in the infrared search and track (IRST) system. The current methods typically exhibit strong performance in simple backgrounds, but they struggle to effectively suppress high-brightness background and strong noise. For the sake of diminish the false alarm rate of infrared small target detection algorithms under strong clutter interference, an infrared small target detection algorithm based on variance difference weighted three-layer local contrast measure (VDWTLCM) is proposed in this paper. The algorithm comprises two modules: the three-layer local contrast measure (TLCM) and the weighting function based on mean value of the variance difference (MVD). In the calculation module of TLCM, the image is processed pixel by pixel using a three-layer sliding window, while both the ratio form definition and the difference form definition are utilized to calculate the local contrast for each pixel, which can enhance the target and eliminate the high-brightness background. To calculate the MVD weighting function, a modified layered gradient kernel is first introduced to pre-process the IR image, aimed at achieving more effective clutter smoothing. And then the concept of variance difference is employed to further suppress the highlighted background edges. Ultimately, the target is extracted utilizing adaptive threshold segmentation. The experimental results demonstrate that, compared with the common infrared small target detection algorithms, the proposed algorithm improves the signal to clutter ratio gain (SCRG) by an average of 13.13 times and the background suppression factor (BSF) by an average of 10.02 times, and can achieve superior detection performance in different complex scenarios.