基于特征显著性融合的红外小目标检测

复杂背景下的红外图像通常存在信噪比低、邻近像素灰度变化不明显以及易被杂波信号和噪声干扰的特点,导致红外小目标检测困难。为解决上述问题,提出一种基于特征显著性融合的红外小目标检测算法。首先,在空间域中利用目标与其局部背景灰度差异来计算得到灰度显著图,在频域中结合谱残差计算得到背景抑制后的频域显著图;其次,将灰度显著图和频域显著图归一化后通过哈达玛乘积相互融合;最后,通过自适应阈值分割并使用Unger滤波器剔除较小的噪声点,从而提取出目标区域。实验结果表明,所提算法对图像的信噪比有了数十倍的提升,对背景抑制效果显著,并有着检测率高和虚警率低的优点,是一种有效的小目标检测算法。

Infrared Small Target Detection Based on Feature Saliency Fusion

Infrared images with complex background are usually characterized by low signal-to-noise ratio (SNR), insignificant grayscale changes of adjacent pixels, and easy interference by clutter signal and noise, which makes detection of small infrared targets difficult. In order to solve the above problems, an infrared small target detection algorithm based on feature saliency fusion was proposed. Firstly, in the spatial domain, the grayscale difference between the target and its local background was used to calculate the grayscale saliency map, and in the frequency domain, the background suppressed frequency domain saliency map was calculated by combining the spectral residuals. Secondly, the grayscale significance map and frequency domain significance map were normalized and fused with each other by Hadamard product. Finally, the target region was extracted by adaptive threshold segmentation and Unger filter to eliminate small noise points. Experimental results show that the proposed algorithm can improve the image SNR by tens of times, and has a significant effect on background suppression, and has the advantages of high detection rate and low false alarm rate, which is an effective small target detection algorithm.