Entropy-based window selection for detecting dim and small infrared targets

Dim and small target detection in complex background is considered a difficult and challenging problem. Conventional algorithms using the local difference/mutation possibly produce high missed or mistaken detection rates. In this paper, we propose an effective algorithm for detecting dim and small infrared targets. In order to synchronously enhance targets and suppress complex background clutters, we adopt an adaptive entropy-based window selection technique to construct a novel local difference measure (LDM) map of an input image, which measures the dissimilarity between the current region and its neighboring ones. In this way, the window size can be adaptively regulated according to local statistical properties. Compared with the original image, the LDM map has less background clutters and noise residual. This guarantees the lower false alarm rates under the same probability of detection. Subsequently, a simple threshold is used to segment the target. More than 600 dim and small infrared target images against different complex and noisy backgrounds were utilized to validate the detection performance of the proposed approach. Extensive experimental results demonstrate that the proposed method not only works more stably for different target movements and signal-to-clutter ratio values, but also has a better performance compared with classical baseline methods. The evaluation results suggest that the proposed method is simple and effective with regard to detection accuracy.