摘要
针对利用各向异性扩散方程的去噪模型在求解中存在计算量大、耗时长、影响实时性等缺点,本文充分利用并行知识,提出了有效的解决方案。即基于各向异性扩散去噪模型,设计工作站机群平台,对噪声图像进行条状重叠的数据划分,以便实现算法节点内与节点间的两级并行策略:在机群结点内部采用共享内存结构,机群节点间采用分布内存结构,以二者的最优结合实现并行的层次结构化,从而得到一种高效的多层次并行图像去噪算法。实验结果表明,在基于混合模型的并行环境下,该算法能在一定程度上提高原算法的计算效率,不仅有效地缩短了运行时间,而且仍能获得与其相当的图像去噪质量。
According to the shortcomings of the anisotropic diffusion equation denoising model such as intensive calculations , time consuming, affecting the real-timeness, etc, a full use of parallelism knowledge is made to put forward an effective solution. Based on the idea of the anisotropic diffusion equation denoising model, we design a cluster of workstations and divide the noise image into overlapping strips to realize the two-level-parallel strategies：the intra-node cluster using shared memory structure, the inte-node cluster using the distributed memory structure, the optimal combination of the two is used to achieve the parallel structure. Finally an effective hierarchical parallel algorithm for denoising images is proposed. The test result shows that , based on the hybrid-model parallel environment, the operating efficiency of the algorithm can be greatly enhanced,and the running time can be greatly reduced, meanwhile the comparable denoising quality can still be obtained.
出处
《计算机工程与科学》
CSCD
北大核心
2010年第4期49-51,66,共4页
Computer Engineering & Science
基金
重庆市科委基金资助项目(CST2005BB0061
KJ070514)
关键词
图像去噪
各向异性扩散方程
机群
并行算法
混合模型
image denoising
anisotropic diffusion equation
cluster
parallel algorithm
hybrid model
