Gaussian Mixture Background Modeling Based on CUDA

摘要：

Pixel-level background modeling is a common computer vision task, and Gaussian mixture modeling (GMM) algorithm is one of the most used methods. Compute Unified Device Architecture (CUDA) is a technology of general-purpose computing on the GPU, which makes users develop GPU program and achieve high-speed parallel computation. In this paper, by harnessing the feature of CUDA, We improved GMM algorithm by parallel processing and memory optimization. Then, the improved GMM algorithm was implemented on GPU. Compared with the traditional approach based on CPU, the experiment results show that the running speed of proposed method has been increased significantly.

关键词： Background Subtraction Gaussian Mixture Modeling GPU CUDA

作者: REN Hao LU Xiao-feng WANG Jia LU Heng-li FAN Tian-xiang

作者单位: School of Communication and Information Engineering, Shanghai University, Shanghai, China

会议类型: 国际会议

会议名称: 2010 International Conference on Circuit and Signal Processing(2010年电路与信号处理国际会议 ICCSP 2010)

会议地点: 上海

会议语种:英文

页码: 307-310

在线出版日期: 2010-12-25（万方平台首次上网日期，不代表论文的发表时间）