Infrared small target detection via multidirectional derivative joint contrast measure

Detecting small targets in infrared imagery plays a crucial role in infrared search and early warning applications, and has recently become a research hotspot within the field of computer vision. However, many existing approaches predominantly rely on single-feature representations, which limits their effectiveness when dealing with complex scenes involving cluttered regions like edges and corners. To overcome this limitation, the present research introduces a novel method for detecting small infrared targets based on a multidirectional derivative joint contrast measure (MDCM). Initially, a facet model is used to derive the multidirectional second-order derivatives (MSOD) of the infrared image, enabling analysis of contrast characteristics between targets and background interference in the MSOD domain. Leveraging the distinct second-order derivative features inherent to small targets, a derivative saliency measure weighted peak difference (DSMWPD) is developed to mitigate the influence of pronounced edges and corner artifacts. Additionally, acknowledging the approximately isotropic nature of small targets, a local contrast measure weighted cross-dissimilarity (LCMWCD) is designed, which applies multidirectional contrast dissimilarities as penalization to suppress intense structured clutter and further emphasize target regions. The final saliency map is obtained by combining the outputs multiplicatively, followed by adaptive thresholding to extract the targets. Experimental validation demonstrates that the proposed approach achieves superior accuracy and robustness in detecting infrared small targets across various challenging background conditions, outperforming several contemporary state-of-the-art algorithms.