Adaptive method for infrared small target detection based on gray-scale morphology and backward cumulative histogram analysis

The small target detection in infrared image sequences is a fundamental step in the process of infrared search and tracking systems. This paper proposes a fast and adaptive method for infrared small target detection using gray-scale morphology and backward cumulative histogram analysis of the image. The proposed algorithm consists of five phases. Firstly, the perceptually insignificant large image background regions are removed using gray-scale morphology processing. Then, we generate backward cumulative histogram of the image after background elimination. To smooth noisy data, the backward cumulative histogram profile is fitted by a low order polynomial. Thirdly, the lower limit of the threshold that best extracts small targets is obtained based on dynamic analysis of the derivative curve of fitted low-order polynomial. Fourthly, optimal threshold is selected based on reasonable range of the threshold and constant false alarm rate. Finally, false targets are further removed by modified multi-frame data association. Experimental results show that our proposed method obtains very high detection rate and extremely low false alarm rate.