An Inpainting Based Infrared Small Target Detection Method

Infrared small target detection technology has an essential role in various target detection field. Nonetheless, it is a challenging task to detect small targets due to the lack of shape, texture, and other information. Recently, many traditional algorithms (e.g., Top-Hat transformation) detect numerous false alarms focusing on local regions of infrared images, which leads they are not sufficient to fulfill the requirements. To address this problem, in this paper, an infrared detection framework is proposed to detect small targets. The framework is composed of Top-Hat transformation and image inpainting. Foremost, Top-Hat with mean filtering detect suspicious target areas. Then, to further refine the targets, a local background estimation is designed into the framework. Moreover, inpainting with diffusion model is employed to obtain the background by the global semantic. Eventually, experimental results on several infrared sequences indicate the proposed framework has an effective detection capability and robustness to different environments.