Infrared Small Target Detection Algorithm Based on Robust Tensor Decomposition Model within Bayesian Framework

Abstract

Small targets detection in infrared video can be further improved by considering that the background has high correlation and low rank characteristics while foreground objects maintain sparsity. In this paper, a new infrared small target detection algorithm within Bayesian framework is proposed. A three-dimensional tensor structure of the video sequence is supposed to be decomposed into low rank background, sparse foreground and noise. The corresponding probabilistic models for the three parts form a Bayesian network which is solved by using variational Bayesian inference. Finally, the isolated sparse component is utilized for further target detecition. Experimental results show that the proposed method is suitable for the detection of small infrared target with good detection accuracy and robustness.