In many applications, such as observation systems, it is important to detect and identify a possible threat as early as possible. Especially the approaching ones are of interest. Such moving threats/objects are often at a very large distance and are observed as very small and unrecognizable objects at the image plane of a camera.
In this research multi-frame super-resolution (SR) reconstruction methods are developed that are capable of improving the resolution, signal-to-noise ratio and contrast of moving objects in under-sampled image sequences. These improvements will help to increase the detection and recognition rate of moving objects of various sizes. The most difficult case is improving the small moving objects which consist of solely “mixed” boundary pixels on a cluttered background. We developed a SR reconstruction method that performs well on both simulated and real-world data and it is shown that, for reconstructing small moving objects, our method outperforms state-of-the-art existing SR reconstruction methods.