English

应用科学学报 >

2014 , Vol. 32 >Issue 1: 85 - 92

DOI: https://doi.org/10.3969/j.issn.0255-8297.2014.01.014

信号与信息处理

非线性图像扩散LB 模型的CUDA 算法设计与实现

展开
  • 上海大学通信与信息工程学院，上海200072
严壮志，教授，博导，研究方向：生物医学图像与信息处理，E-mail：zzyan@shu.edu.cn

收稿日期: 2013-01-29

  修回日期: 2013-03-25

  网络出版日期: 2013-03-25

基金资助

国家自然科学基金(No.61171146)；上海市科委科技创新行动计划基金(No.11DZ1921702)资助

收起

Design and Implementation of CUDA Algorithms Based on Nonlinear Image Diffusion LB model

Expand
  • School of Communication and Information Engineering, Shanghai University, Shanghai 200072, China

Received date: 2013-01-29

  Revised date: 2013-03-25

  Online published: 2013-03-25

Fold

摘要

为提高基于格子波尔兹曼(Lattice Boltzmann, LB) 模型图像去噪方法的性能，研究了非线性图像扩散LB 模型的CUDA 算法，即分别利用纹理内存、共享内存以及直接使用全局内存来实现非线性图像扩散LB 模型中的迁移过程. 利用合成图像和真实图像的去噪实验表明，针对非线性图像扩散LB 模型，GPU 相对CPU 的加
速比可达90 倍以上；而且加速比的提高与GPU 流处理器的数目成正比.

关键词： 图像去噪; 非线性图像扩散; LB 模型; CUDA 算法

本文引用格式

周明， 严壮志， 黄彬 . 非线性图像扩散LB 模型的CUDA 算法设计与实现[J]. 应用科学学报, 2014 , 32(1) : 85 -92 . DOI: 10.3969/j.issn.0255-8297.2014.01.014

Abstract

 To improve the performance of Lattice Boltzmann (LB) method in image denoising, this paper proposes three compute unified device architecture (CUDA) algorithms to realize streaming processes of LB in nonlinear image diffusion, which are based on texture memory, shared memory, and global memory, respectively.To test effectiveness and efficiency of the GUDA algorithms, experiments were carried out with natural and composite images. The results show that GPU acceleration is 90 times faster than CPU acceleration, and the factor is proportional to the number of GPU stream processors.

Key words：  image denoising; nonlinear image diffusion; LB model; CUDA algorithm

参考文献

[1] JAWERTH B, LIN P, SINZINGER E. Lattice Boltzmann models for anisotropic diffusion of images [J]. Journal of Mathematical Imaging and Vision, 1999, 11(3): 231-237.

[2] 陈玉，严壮志，钱跃竑. 基于格子波尔兹曼模型的图像去噪 [J]. 电子学报，2009, 37(3): 574-580.

CHEN Yu, YAN Zhuangzhi, QIAN Yuehong. The Lattice Boltzmann method based image denoising [J]. Acta Electronica Sinica, 2009, 37(3): 574-580. (in Chinese)

[3] CHANG Q S, YANG T. A Lattice Boltzmann method for image denoising [J]. IEEE Transactions on Image Processing, 2009, 12(18): 2797-2802.

[4] ZHANG W H, SHI B C. Application of Lattice Boltzmann method to image filtering [J]. Journal of Mathematical Imaging and Vision, 2012, 43: 135-142.

[5] 王志强，严壮志，钱跃竑. 图像非线性扩散去噪的格子波尔兹曼方法 [J]. 应用科学学报，2010, 28(4): 367-373.

WANG Zhiqiang, YAN Zhuangzhi, QIAN Yuehong. Nonlinear diffusion for image denoising using Lattice Boltzmann method [J]. Journal of Applied Sciences, 2010, 28(4): 367-373. (in Chinese)

[6] ZHAO Y. Lattice Boltzmann based PDE solver on the GPU [J]. The Visual Computer, 2008, 24: 323-333.

[7] WANG Zhiqiang, YAN Zhuangzhi, CHEN G. Lattice Boltzmann method of active contour for image segmentation [C]//Sixth International Conference in Image and Graphics (ICIG), 2011: 338-343.

[8] 张蕊，严壮志，刘玮. 图像修复的格子波尔兹曼方法 [J]. 电子测量技术，2011, 34(3): 46-65.

ZHANG Rui, YAN Zhuangzhi, LIU Wei. Lattice Boltzmann method based image inpainting [J]. Electronic Measurement Technology, 2011, 34(3): 46-65. (in Chinese)

[9] KUZNIK F, OBRECHT C, RUSAOUEN G, et al. LBM based flow simulation using GPU computing processor [J]. Computers & Mathematics with Applications, 59: 2380-2392.

[10] 朱红斌，刘学慧， 柳有权, 等. 基于Lattice Boltzmann模型的液混合流模拟 [J]. 计算机学报，2006, 29: 2071-2079.

Zhu Hongbin, Liu Xuehui, Liu Youquan, et al. Binary Mixtures Simulation Based on Lattice Boltzmann Method [J]. Chinese Journal of computers, 2006, 29: 2071-2079. (in Chinese)

[11] 柳有权，刘学慧，吴恩华. 基于GPU带有复杂边界的三维实时流体模拟 [J]. 软件学报，2006, 17: 568-576.

LIU Youquan, LIU Xuehui, WU Enhua. Real-time 3D fluid simulation on GPU with complex obstacles [J]. Journal of Software, 2006, 17: 568-576. (in Chinese)

[12] ZHOU H, MO G Y, WU F, et al. GPU implementation of Lattice Boltzmann method for ?ows with curved boundaries [J]. Computer Methods in Applied Mechanics and Engineering, 2012, 225-228: 65-73.

[13] OBRECHT C, KUZNIK F, TOURANCHEAU B, et al. A new approach to the Lattice Boltzmann method for graphics processing units [J]. Computers & Mathematics with Applications, 2011, 61: 3628-363.

[14] TOLKE J. Implementation of a Lattice Boltzmann kernel using the compute unified device architecture developed by Nnidia [J]. Computing and Visualization in Science, 2010, 13: 29-39.

[15] 陈曙晖，熊淑华. 大规模并行处理编程实战 [M]. 北京：清华大学出版社，2010: 106-116.

CHEN Shuhui, XIONG Shuhua. Programming massively parallel processors: a hands-on approach [M]. Beijing: Tsinghua University Press, 2010: 106-116. (in Chinese)

[16] http://www.realworldtech.com/amd-nvidia-gpu-performance/.

[17] http://www.realworldtech.com/gpu-memory-bandwidth/.
Options
文章导航

/

〈 〉