Shearlet-Based Structure-Aware Filtering for Hyperspectral and LiDAR Data Classification

Overall accuracy vs. the initial number of pixels inside superpixel () and the number of regions in the final merging map () with different numbers of training samples per class on the Houston dataset (first row), Trento dataset (second row) and MUUFL Gulfport dataset (third row). The number of training samples per class is 3 (first column), 5 (second column), 10 (third column), and 15 (last column).