基于自适应SUSAN各向异性扩散的红外弱小目标检测

Infrared Small Target Detection Based on Adaptive SUSAN-controlled Anisotropic Diffusion

针对复杂背景下红外弱小目标提取困难的问题,提出了一种自适应SUSAN各向异性扩散的红外弱小目标检测算法。该算法结合SUSAN边缘检测算子与各向异性扩散,形成新的扩散方程对红外图像进行背景预测,与原图像差分后实现弱小目标检测。为使算法具备自适应能力,提出SUSAN边缘检测器灰度差阈值的自适应设定方法,采用绝对偏差中值算子作为其扩散系数。实验结果表明,该算法能够有效滤除复杂图像背景,大幅提升信噪比,同时保留目标大小。

To solve the difficult problem of acquiring the small target from infrared image, a new algorithm based on adaptive SUSAN-controlled Anisotropic Diffusion is proposed. The new algorithm combines the SUSAN edges detection algorithm and anisotropic diffusion so that a new diffusion equation is formed, which is used for predicting the background of the infrared image. After that the small targets are extracted from the residual picture between the original image and background image. To make an improvement of the adaptive ability of the new algorithm, a new way to set the limen of the SUSAN edge detector is proposed, and the median absolute deviation is used as the diffusion coefficients. The experiment demonstrates that the proposed algorithm is able to suppress the background of the image effectively, improve the SNR obviously and preserve the size of target accurately.