In numerous prior works, a technique called Artificial Color has been developed to extract pixels belonging to a prespecified class while rejecting pixels belonging to other prespecified sets with great reliability. The heart of the algorithm is another well described pattern recognition method called Margin Setting. Margin Setting achieves highly reliable classification by refusing to classify some borderline pixels. As a result, the image produced using Artificial Color methods is reliable in finding the target of interest but that target may contain unclassified pixels, leading to a spotty or ragged image being extracted. It is showed here that post processing that ragged image using mathematical morphology can improve the extracted image substantially, and median filtering after that produces even more improvement.
Jian Fu, H. Caulfield, Alak Bandyopadhyay, "Pairing Mathematical Morphology with Artificial Color to Extract Targets from Clutter" in Journal of Imaging Science and Technology, 2007, pp 148 - 154, https://doi.org/10.2352/J.ImagingSci.Technol.(2007)51:2(148)