大规模虚拟地形的图像径向权混合生成算法  

Algorithm for Synthesizing Large-scale Virtual Terrain from Images Using Radially Weighted Blending

作　　者：庞明勇[1,2] 赵瑞斌[1,2] 

机构地区：[1]南京师范大学教育技术系,江苏南京210097 [2]北京大学视觉与听觉信息处理国家重点实验室,北京100871

出　　处：《南京理工大学学报》2009年第4期438-443,共6页Journal of Nanjing University of Science and Technology

基　　金：国家自然科学基金(60873175);江苏省高校自然科学基础研究项目(07KJD460108);北京大学视觉与听觉信息处理国家重点实验室开放课题基金

摘　　要：结合图像处理技术和径向权混合方法,提出一种基于图像集的大规模虚拟地形自动生成算法。该算法首先由多幅图像色彩数据生成标准化地形拼接块模型,并通过滤波处理消除地形块内部的跳变现象;借助定义在地形块上的有限支撑径向权函数,对毗邻地形块的子块进行加权混合处理,实现了地形块的无缝拼接平滑过渡。该算法支持相关参数的交互式调整,使用户能够对地形的形状和平滑程度等进行有效控制。实验结果表明:该算法可以自动、快速、高效地生成满足不同形状和规模要求的虚拟地形模型。A novel algorithm to synthesize large-scale virtual terrains from a set of images is presen- ted based on digital image processing technologies and radially weighted blending method. The algorithm first creates standardized terrain-blocks from the images, and smoothes the blocks by employing fast Fourier transform and Butterworth＇s low-pass filter. A radial weight defined on each block, which is utilized 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 operators with the radial weights. The algorithm also provides users a method to control the terrains effectively by adjusting several control parameters. Experimental results show that the algorithm can generate terrains with various scales and styles automatically, fast and effectively.

关 键 词：地形合成 图像融合 虚拟现实 数字高程模型 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]