Infrared small target detection based on spatio-temporal saliency in video sequence

Infrared small target detection plays an important role in military field. In this paper, we propose a new method to detect infrared small target in video sequence based on spatio-temporal saliency using phase spectrum of quaternion fourier transform. Four second-order directional derivative maps at four different directions are used to create two feature channels and the intensity and motion information are added as other two feature channels. All feature channels construct a quaternion. Then, the quaternion fourier transform is calculated to acquire the phase spectrum. Finally, the small target detection map is obtained through inversing QFT only using the phase spectrum with gaussian filter. The experimental results demonstrate that the proposed method has better performance than other four state-of-art methods for the infrared video sequences.