A novel infrared dim small target detection algorithm based on frequency domain saliency

摘要

Infrared small target detection technology has great significance in the infrared guidance, infrared warming, video surveillance and other fields. From the perspective of human visual saliency, our paper proposes a novel infrared dim small target detection algorithm based on frequency domain saliency. First, we get saliency map in the frequency domain. The algorithm combines the amplitude spectrum filtered by Gaussian low-pass filter with an appropriate scale and the phase spectrum, and then obtain saliency map by inverse Fourier transform. Secondly, we get saliency regions by segmenting the saliency map, and then use the gray centroid method to locate saliency regions. Finally, the small target is identified by local contrast algorithm. Experimental results show that our algorithm proposed higher detection accuracy for infrared images with different background, and can also suppress global repetitive clutter. Moreover, the robustness of our algorithm for noise is stronger.