摘要
为解决传统算法在高分辨率图像局部模糊检测方面处理速度较慢的问题,文章提出一种面向GPU的图像局部模糊检测的并行加速方案。利用GPU的强大计算能力加速局部模糊检测中均方差的计算、再模糊处理、均方差比较和清晰区域标记等过程。结果表明,基于GPU的图像局部模糊检测并行算法的性能与基于CPU的串行算法相比,可获得270倍的加速比,能够为大规模实时性图像处理系统的应用设计提供参考。
In order to solve the problem of slow processing speed of traditional algorithms in high-resolution image local blur detection,a GPU-based parallel acceleration scheme for image local blur detection is proposed in this paper. The powerful computing power of GPU is used to accelerate the local-blur detection procedures such as the calculation of mean square error, re-blurring, comparison of mean square error and labeling clear region. The results show that the performance of the GPU-based parallel algorithm for image local blur detection can achieve a speedup of 270 times compared with the CPU-based sequential algorithm, which can provide reference for the application design of large-scale real-time image processing systems.
出处
《大众科技》
2023年第1期9-13,共5页
Popular Science & Technology
基金
广西高校中青年教师科研基础能力提升项目(2019KY0675)
广西自然科学基金项目(2021JJB170060)
广西创新驱动发展专项资金项目(桂科AA18118036)
梧州学院校级科研项目(2022B006)
梧州学院2022年自治区大学生创新创业训练计划立项项目(S202211354103)。
关键词
局部模糊
并行加速
再模糊
均方差
local blur
parallel acceleration
re-blurring
mean square error
