【摘要】 为提高图像处理领域协方差矩阵的计算效率,满足其在实时要求下的应用,借助GPU通用计算技术,结合CUDA编程模型,对协方差矩阵的计算进行有针对性的并行化优化,设计并实现一种高效的并行图像协方差矩阵算法。为在通用PC平台上使用协方差矩阵并满足实时性需求的各种图像处理应用提供了一个可行的解决方法,对其它领域涉及到协方差矩阵的实时计算也有良好的借鉴作用。与原有的CPU实现方法相比,GPU的效率有了平均数千倍的提升。更多还原

【Abstract】 To improve the efficiency of covariance matrix computation for image processing to adapt to real-time or near real-time requirements of practical applications,with the general purpose of GPU technology,the covariance matrix computation was optimized for parallel computation and a novel parallelization approach for covariance matrix computation based on CUDA programming model was proposed.Image processing applications based on covariance matrices can be processed in real time on general personal PC platforms.Moreover,as to image processing,there are many other real-time requirements based on covariance matrices can also be satisfied.Compared to the original CPU implementation,the proposed GPU implementation improves the efficiency by thousands of times on average.更多还原