Data of Infrared Small Target Detection Using Local Component Uncertainty Measure With Consistency Assessment

The development of effective detection algorithms under a complex background for small infrared (IR) targets has always been difficult. The existing algorithms have poor resistance to complex backgrounds, easily leading to false alarms. Furthermore, each target and its background correspond to different component signals, and changes in the components in space cause observation uncertainty. Inspired by this phenomenon, we propose a method for detecting small targets in complex backgrounds using local uncertainty measurements based on the compositional consistency principle. First, a multilayer nested sliding window is constructed, and a local component uncertainty measure algorithm is used to suppress the complex background by evaluating the component comprising local area signals. Subsequently, an energy weighting factor is introduced to reinforce the energy information embedded in the target in the uncertainty distribution map, thereby enhancing the target signal. Validation results obtained on real IR images show that the energy-weighted local uncertainty measure performs better when detecting small targets hidden in complex backgrounds, with a high signal-to-clutter ratio (SCR) gain and background suppression factor (BSF). The effectiveness of our proposed method on several typical open-source datasets is provided in this dataset, and quantitatively compared with several other sets of state-of-the-art algorithms.