This is the authors' abstract. We don't add key points for this paper.
Superresolution mapping (SRM) based on the Hopfield neural network (HNN) is a technique that produces land cover maps with a finer spatial resolution than the input land cover fraction images. In HNN-based SRM, it is assumed that the spatial dependence of land cover classes is homogeneous. HNN-based SRM uses an isotropic spatial dependence model and gives equal weights to neighboring subpixels in the neighborhood system. However, the spatial dependence directions of different land cover classes are discarded. In this letter, a revised HNN-based SRM with anisotropic spatial dependence model (HNNA) is proposed. The Sobel operator is applied to detect the gradient magnitude and direction of each fraction image at each coarse-resolution pixel. The gradient direction is used to determine the direction of subpixel spatial dependence. The gradient magnitude is used to determine the weights of neighboring subpixels in the neighborhood system. The HNNA was examined on synthetic images with artificial shapes, a synthetic IKONOS image, and a real Landsat multispectral image. Results showed that the HNNA can generate more accurate superresolution maps than a traditional HNN model.
Answers are grounded in this paper and cite their sources.
No takes yet. Share an insight, caveat, or question.
Li et al. (2014) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: