A Super-Resolution Mapping Algorithm Based on the Level Set Method

The presence of mixed pixels is a recurring problem in extracting accurate land cover information from remote sensing images. To deal with the mixed-pixel problem, we propose to find the land cover map at the resolution higher than the observed remote sensing image. The process to obtain this higher resolution land cover map is called "super resolution land cover mapping (SRLCM)." In this work, we modeled the problem of the SRLCM as an image segmentation problem where the level set method can be applied to find the boarders between land cover classes at the sub-pixel accuracy. Our experimental results show that our proposed approach can significantly improve the classification accuracy over the Maximum likelihood classifier.