Synthesizing Large-Scale Virtual Terrain from Image Atlas
Ruibin Zhao, Mingyong Pang · 2008
In this paper, a novel algorithm to synthesize large-scale virtual terrain from a set of images is presented based on digital image processing technologies and radial weighted blending method. The algorithm first creates standardized terrain-blocks from the images, and then smoothes the blocks by employing fast Fourier transform (FFT) and butter worth low-pass filter. A radial-weighted defined on each block, which is used to synthesize four local neighboring blocks to a single blended area, is subsequently constructed for each block. As a result, a complete and smooth large-scale terrain can be obtained by blending operator with the radial-weights. The algorithm also provides user a tool of controlling terrain by adjusting several parameters. Experiments show that our algorithm can generate terrains with various scales and styles, fast and effectively.