空间自适应卷积核滤波红外弱小目标检测

Space-adaptive Convolution Kernel Filtering For Infrared Dim Target Detection

为了减少复杂红外图像中平滑背景边缘的影响,将具有各向异性特性的PM扩散模型应用到红外弱小目标检测,提出了空间自适应卷积核滤波检测算法,并对扩散系数进行了优化。针对模型中扩散参数难以确定的问题,提出了一种利用Sobel边缘检测算子估计扩散参数的方法。滤波后采用信噪比(SNR,Signal Noise Ratio)和接受机工作特性(ROC,Receiver Operating Characteristic)曲线进行性能评价,实验结果表明,与PM扩散模型滤波和中值滤波相比,该算法有效抑制了边缘,大大提高了信噪比,提高了检测概率,降低了虚警概率,具有更好的性能。

In order to reduce the effect that caused by smooth background edge of the complex infrared image, space-adaptive convolution kernel filtering operator is proposed for infrared dim target detection based on the anisotropy PM diffusion model. And the diffusion parameter is optimized. A method of diffusion parameterestimation based on Sobel edge detection operator is proposed. SNR and ROC curve are used to evaluate the method. The experimental results show that the edge is suppressed effectively, the SNR is improved greatly, the detection probability is improved, and the false-alarm probability is reduced compared with the PM diffusion model filtering and median filtering.