Patch-based Monte Carlo Terrain Upsampling via Gaussian Laplacian Pyramids
Wanwan Li · 2023
Terrain upsampling techniques are aiming at solving the changing problem of synthesizing arbitrary dimensional high-resolution terrain heightmaps from given low-resolution terrain heightmaps. In this paper, we propose a novel patch-based Monte Carlo terrain upsampling approach using Gaussian Laplacian pyramids. In our approach, given low-resolution terrain heightmaps as input, we generate its Gaussian Laplacian pyramids. Then, we upsample the Laplacian pyramids using a Monte Carlo image patch quilting approach. Finally, we reconstruct realistic terrains from the upsampled terrain Laplacian pyramids. In the end, We conduct a series of numerical experiments to show the effectiveness of our approach in upsampling terrain heightmaps.