Infrared small target detection with directional difference of Gaussian filter

Robust detection of infrared dim and small target is still a challenge for the complexity of background clutter. In previous work, the neighboring pixels around target are often processed simultaneously to enhance target or suppress background, ignoring the directional difference of nearby background. This results in the difficulty of distinguishing real targets and background edges, and causes a high false alarm rate. In this paper, the directional difference of Gaussian (DDoG) filter is proposed, which has a fan-like shape and is sensitive to orientations, then can reduce the interference of background edges. Extensive experiments verify the feasibility and effectiveness of our proposed method comparing with 4 state-of-art methods.