Vol. 54, Issue 3, pp. 365-381 (2024)

Abstract

To solve the problem of false alarm rate in detecting infrared small targets under complex cloud backgrounds, a novel algorithm combining structure tensor and local contrast is proposed. The structure tensor can better describe the gradient distributions in the local image area, and its eigenvalues can also depict the characteristics of the area. Combining the weighted local contrast with eigenvalues, the small targets can be enhanced and the background can be suppressed. In addition, to highlight the target, the regional complexity is further used for weighting local contrast. The presented algorithm steps are as follows: firstly, Gaussian filtering is performed on the original image; secondly, the larger eigenvalue of the structure tensor matrix is used to calculate the local contrast through the difference operation; thirdly, the regional complexity is calculated by the gray difference between the central and surrounding regions for weighting the local contrast to generate a saliency map; finally, an adaptive threshold segmentation is performed on the saliency map to extract the real target. The comparative experiments show that the proposed algorithm can achieve the highest detection rate, lowest false alarm rate, and shortest running time.