Entropy flow-based approach to detect small target in the sea-sky background

In this paper, we introduce a bionic entropy flow-based approach to the detection of small target in the sea-sky background. Entropy flow is a novel descriptor of image motion, and a differential method is applied to compute its flow field. The detection of small moving target is divided three steps: firstly, compensate the background motion through 2-dimensional histogram of entropy flow; secondly, use Robinson filter to eliminate spurious motions; thirdly, a modified local entropy-based method to segment the target. The experiments suggest that the algorithm proposed in this paper provides instructive viewpoint for small targets detection.