基于小波变换和GPU加速的地形绘制技术研究

Real-time Rendering Technology for Terrain Based on Wavelet Translation and GPU Acceleration

研究基于GPU和小波变换的海量地形数据实时绘制技术,通过小波变换构造地形网格的动态多分辨率模型,利用小波完备的数据重建功能,实现不同分辨率数据间的实时转换,减少从慢速外存向内存读数据的时间开销,提高绘制效率。通过小波渐近绘制算法解决不同分辨率模型间的平滑过渡问题,通过扩大了小波变换的采样范围解决块间小波变换的边界问题,通过顶点数据拟合算法解决相邻地形块间的裂缝问题,优化GPU的地形遮挡剔除算法,挖掘GPU的通用计算能力,在加速几何绘制的同时加速小波变换,进一步提高算法的效率。

An approach of realtime rendering large-scale terrain based on GPU acceleration and wavelet translation was proposed.It firstly constructs dynamic multi-resolution models of terrain grid through wavelet translation,which can reach realtime switching between different resolution models via data reconstructing so as to reduce the cost of loading data from slow out-of-core storage to main memory.Then it realizes smooth transition between different resolution models via wavelet-progressive-reconstruction algorithm,removes boundary distortion of wavelet translation through expanding original sampling range,stitches cracks in neighboring terrain blocks through vertex data fitting algorithm.Finally,it improves rendering efficiency through optimizing terrain occlusion culling algorithm based on GPU and accelerating wavelet translation based on GPU.