摘要
纹理绘制技术通过纹理线条和颜色变化能够细致且生动地表现2D矢量场的速度、方向以及数据相关性等特征信息,但扩展到3D矢量场空间时,由于3D矢量场本身的空间特性容易造成纹理单元之间产生严重的视线遮挡问题,影响研究人员对矢量场内部固有属性特征的观察和分析.针对此问题,提出一种基于GPU加速实现的稀疏噪声纹理生成的改进3D矢量场Volume LIC绘制技术.在噪声生成部分,基于泊松盘分布以避免噪声点间的相互遮挡,采用Hilbert空间填充线遍历减少生成噪声点的规律性和人工痕迹,并通过高斯滤波核滤除高频区域生成稀疏高斯噪声.整个算法采用GPU+GLSL硬件加速机制,在噪声纹理采样时,利用GPU顶点颜色线性插值功能和片元计算方法有效地加速LIC纹理生成过程,并将卷积噪声和矢量场数据作为纹理传入GPU;采用光线投射算法实现LIC纹理的3D绘制显示,并通过光线提前终止技术和空白空间跳跃技术有效提升绘制效率;同时提供多种有效的交互分析手段查看流场内部特征.实验结果表明,该方法生成的3D纹理图像清晰、绘制效率高,能够有效地缓解3D复杂矢量场卷积数据过多引起的遮挡与混乱现象,具备良好的可视化效果.
Texture rendering technology can meticulously and vividly visualize characteristic information of 2Dvector field such as speed, direction and data correlation by texture lines and color variations. But when extendedto 3D flow field, it can’t help researchers to find and analyze the inherent attributes of flow filed characteristics asthe problem of occlusion and confusion existed between texture slices caused by spatial characteristic of 3D vectorfield. To solve this problem, a new sparse noise based improved VolumeLIC rendering technology of 3D vectorfield implemented by GPU acceleration is proposed in this paper. Noise generation followed by Poisson-diskdistribution can help noise points to avoid shading each other. To decrease regular patterns and undesired artifactsin the final sparse noise texture, cube-filling Hilbert curve is adopted to traverse the points. And Gaussian filterkernel is adopted to filter out the high frequency region and generate sparse Gaussian noise. The overall algorithmis implemented by GPU hardware acceleration mechanism, GPU based vertex color linear interpolation functionand fragment calculation method are used in nose texture sampling step to accelerate the process of LIC texturegeneration. Finally the LIC noise texture and flow field data are introduced into GPU, GPU based ray casting algorithmis adopted to visualize LIC texture. In the period of ray casting, early-ray termination and empty-spaceskipping are utilized to effectively improve the rendering efficiency. Meanwhile a variety of effective interactiveanalysis methods are introduced to check the internal flow field characteristics. Experimental results show that the3D texture image generated by our method is distinct and delicate, our method can improve the rendering speed,effectively alleviate the problem of occlusion and confusion occurred by too much convolution data in the 3Dcomplex flow field and has nicer effect.
出处
《计算机辅助设计与图形学学报》
EI
CSCD
北大核心
2016年第5期723-732,共10页
Journal of Computer-Aided Design & Computer Graphics
基金
国家自然科学基金(61303127)
四川省科技厅项目(2014SZ0223
2014GZ0100
2015GZ0212)
