海量地形实时动态存储与绘制的GPU实现算法

摘要

为了降低实时更新和存储海量地形的形变数据对动态绘制速度的影响,提出一种基于整数小波变换与限制性四叉树相结合的GPU并行动态存储与绘制算法.首先设计面向CUDA并行且无损的基于块的整数小波变换算法和SPIHT压缩算法,提高地形压缩比以减小数据传输量,同时解决了海量地形动态数据存储的编解码的实时性问题,实现了局部动态地形数据的实时存储；然后将小波系数、限制性四叉树层次结构以及模板技术相结合,提出一种自适应三角化和绘制的并行处理算法.实验结果表明,对于海量地形数据,文中算法可以在实现后端及时保存局部形变数据的同时,前端可以保持较高的绘制帧率.%To alleviate the impact on dynamic rendering massive terrain caused by real-time updating and storing deformation data,a GPU friendly parallel rendering algorithm based on integer wavelet transformation and restricted quad-tree is proposed.By designing block-wise parallel integer wavelet transformation and SPIHT algorithms with CUDA,the high compression ratio is achieved and the transmission of massive terrain data is reduced effectively.Meanwhile,the real-time terrain encoding and decoding caused by dynamically changing the terrain data is addressed also.As a result,the realtime storage of locally dynamic terrain data is achieved.Then,a parallel adaptive triangulation and rendering algorithm is proposed by the combination of integer wavelet coefficients,restricted quad-tree and template technologies.Experimental results show that the proposed algorithm can achieve high rendering frame rate for massive terrain data in the front-end,while the local deformation of terrain data due to external forces can also be saved timely in the back-end.