Small infrared target detection based on spatio-temporal fusion saliency

Abstract

A new algorithm for small infrared target detection is proposed by analyzing time domain and spatial domain of infrared image and a model combining spatio-temporal saliency is put forward. Spatio-temporal features are spectral residual from spatial domain and movement information extracted from temporal domain by improved optical flow method combined with frame difference method. After extracting these features, salient regions as the regions of interest can be obtained. The experiment shows, our algorithm has good performance in small infrared target. The recall and precision rate of our algorithm can reach 95.43% and 76.07% This method also has high adaptability and it can lay a good foundation for the following study like tracking in this field.