Answers are grounded in this paper and cite their sources.
This paper proposes a novel class allocation strategy in units of object (UOO) for soft-then-hard super-resolution mapping (STHSRM). STHSRM involves two processes: 1) subpixel sharpening and 2) class allocation. The UOO is implemented in the second process by integrating the object boundaries as an additional structural constraint. First, UOO obtains the object boundaries from remote-sensing images by image segmentation. The number of subpixels within an object is then calculated for each class to meet the coherence constraint of fractional images imposed by soft classification. Finally, a linear optimization model is built for each object to obtain the optimal hard class labels of subpixels. A synthetic image and two real remote-sensing images are used to evaluate the effectiveness of UOO. The results are compared visually and quantitatively with two existing class allocation methods: 1) the highest attribute values first (HAVF) and 2) units of class (UOC). The experimental results show that UOO performs better than these two methods. UOO can better reduce the salt and pepper effect in resultant maps than both HAVF and UOC when dealing with real remote-sensing images. Moreover, UOO can better maintain the structure of land-cover patches, with smoother boundaries as compared with the two methods.
No takes yet. Share an insight, caveat, or question.
Chen et al. (2016) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: