为了减少红外图像中背景边缘对检测的影响, 提出了一种具有鲁棒性的弱小目标检测算法, 该算法利用核各向异性扩散模型进行背景预测, 再与原图像差分实现弱小目标检测。为了提高算法的自适应能力, 提出了一种鲁棒性扩散系数, 能够根据图像背景的起伏程度自适应调整扩散系数曲线的陡峭程度。实验结果表明, 与现有的检测算法相比, 该算法能够在不同类型的复杂背景下有效抑制背景及其边缘, 保留目标大小, 降低虚警率, 具有更强的鲁棒性。

In order to reduce the effect that caused by background edge of the infrared image, a robust small target detection algorithm is proposed. Firstly, a kernel anisotropic diffusion model is used to predict the background. Then, targets are extracted from the residual map between the original image and its predicted background image. In order to improve the algorithm’s adaptive ability, a robust diffusion parameter is proposed which can automatically adjust diffusion parameter based on the fluctuation of the image’s background. The experimental results demonstrate that the proposed algorithm has better performance in the suppression of background and its edge, the reservation of target’s size, the reduction of false-alarm probability, and the robustness compared with the state-of-the-art algorithms under various typical complex backgrounds.