基于CUDA的快速视频流超分辨率重建

李映; 胡杰; 李海良; 申强

0
0
浏览
下载

摘要
关键词
基金信息
论文图表
同行评议
相关论文
评论

Fast Video Stream Super Resolution Reconstruction based on CUDA

首发时间：2015-12-11

LI Ying ¹
Ying Li received the B.S. degree from Northwestern Polytechnical University, Xi’an, China in 1991, the M.S degree from Airforce Engineering University, Xi’an, China in 1999, and the Ph. D degree from the Xidian University, Xi’an, China in 2002. She was a postdoctoral researcher with the School of Computer Science, Northwestern Polytechnical University, Xi’an, China from Feb. 2003 to Apr. 2005. Currently, she is a professor at the School of Computer Science, Northwestern Polytechnical University. From Jan. 2009 to Jul. 2009, she was a visiting researcher in the School of Information Technologies, University of Sydney, Australia. Her interests include SAR image processing, harmonic analysis, computation intelligence, and signal processing.
HU Jie ¹
Hu Jie was born in 1987. He received the M.S. degree in computer science and technology from Northwestern Polytechnical University, Xi'an, China, in 2011. He is currently pursuing his Ph.D degree in the School of Computer Science, Northwestern Polytechnical University, Xi'an, China. His interests include image/video super resolution reconstruction, signal sparse representation, machine learning and high performance computing.
LI Hailiang ² SHEN Qiang ³

1、School of Computer Science, Northwestern Polytechnical University, Xi'an, China, 710129
2、Department of Electronic and Information Engineering, the Hong Kong Polytechnic University, Hong Kong, China
3、Department of Computer Science, Aberystwyth University, UK

Abstract：This paper presents a parallel GPU-based solution for video stream super resolution reconstruction. We propose an approach, using the computer unified device architecture (CUDA) platform developed by NVIDIA, to partition the steps of the non-local iterative back projection (NLIBP) algorithm (which is designed for single image super resolution reconstruction). The approach also exploits the redundant information of the video stream in the time-space domain in an effort to further reduce the unnecessary searching work in the motion estimation process. The use of CUDA enhances the programmability and flexibility for general-purpose computation of GPU. Experimental results show that, with the assistance of CUDA, the processing time is approximately 8 times faster than that of using CPU only in C++ language, while preserving good visual quality of the reconstructed video stream.

keywords： technology of computer application super resolution non-local similarity motion estimation iterative back projction GPU CUDA

点击查看论文中文信息

基于CUDA的快速视频流超分辨率重建

李映 ¹
Ying Li received the B.S. degree from Northwestern Polytechnical University, Xi’an, China in 1991, the M.S degree from Airforce Engineering University, Xi’an, China in 1999, and the Ph. D degree from the Xidian University, Xi’an, China in 2002. She was a postdoctoral researcher with the School of Computer Science, Northwestern Polytechnical University, Xi’an, China from Feb. 2003 to Apr. 2005. Currently, she is a professor at the School of Computer Science, Northwestern Polytechnical University. From Jan. 2009 to Jul. 2009, she was a visiting researcher in the School of Information Technologies, University of Sydney, Australia. Her interests include SAR image processing, harmonic analysis, computation intelligence, and signal processing.
胡杰 ¹
Hu Jie was born in 1987. He received the M.S. degree in computer science and technology from Northwestern Polytechnical University, Xi'an, China, in 2011. He is currently pursuing his Ph.D degree in the School of Computer Science, Northwestern Polytechnical University, Xi'an, China. His interests include image/video super resolution reconstruction, signal sparse representation, machine learning and high performance computing.
李海良 ² 申强 ³

1、西北工业大学，计算机学院，710129
2、香港理工大学，电子与资讯工程学系
3、阿伯里斯特威斯大学，计算机科学系

摘要：本文针对视频流超分辨率重建问题提出了一种基于GPU的并行解决方案。该方法在计算统一设备架构（CUDA）平台上对基于非局部相似性的迭代反向投影这一单针图像超分辨率重建算法进行并行设计与实现。同时该方法提取视频流在时间轴上的大量运动冗余信息，并依据该冗余信息进一步大大减少了算法在运动估计过程中的计算量。CUDA的使用增强了通用GPU计算的可编程性和灵活性。实验结果表明通过使用CUDA，在保持清晰高分辨率重建视频流的同时，相比仅用C++编程语言实现的CPU串行算法，该方法的执行速度能够获得将近8倍的加速比。

关键词：计算机应用技术; 超分辨率; 非局部相似; 运动估计; 迭代反向投影; GPU; CUDA

基金：

论文图表：

引用

导出参考文献

.txt

.ris

.doc

LI Ying,HU Jie,LI Hailiang,et al. Fast Video Stream Super Resolution Reconstruction based on CUDA[EB/OL]. Beijing:Sciencepaper Online[2015-12-11]. https://www.paper.edu.cn/releasepaper/content/201512-675.

No.4663992110995214****

同行评议

共计0人参与

全部评论

0/1000

论文编号	201512-675
论文题目	基于CUDA的快速视频流超分辨率重建
文献类型
收录期刊	上传封面中文期刊英文期刊期刊名称（中文）期刊名称（英文）年，卷（）上传封面中文专著英文专著书名（中文）书名（英文）出版地出版社出版年上传封面中文译著英文译著书名（中文）书名（英文）出版地出版社出版年上传封面中文论文集英文论文集编者.论文集名称（中文） [c]. 出版地出版社出版年， - 编者.论文集名称（英文） [c]. 出版地出版社出版年，- 上传封面中文文献英文文献期刊名称（中文）期刊名称（英文）日期-- 在线地址http:// 上传封面中文文献英文文献文题（中文）文题（英文）出版地出版社,出版日期-- 上传封面中文文献英文文献文题（中文）文题（英文）出版地出版社,出版日期--
英文作者写法：中外文作者均姓前名后，姓大写，名的第一个字母大写，姓全称写出，名可只写第一个字母，其后不加实心圆点“.”, 作者之间用逗号“，”分隔，最后为实心圆点“.”, 示例1：原姓名写法：Albert Einstein,编入参考文献时写法：Einstein A. 示例2：原姓名写法：李时珍；编入参考文献时写法：LI S Z. 示例3：YELLAND R L,JONES S C,EASTON K S,et al.