高通量图像视频计算
发布时间：2018-11-04 21:33
【摘要】：互联网上的图像和视频数据正在飞速地产生和传播.这些数据不仅规模庞大,还具有高并发、高维度、大流量的显著特性,导致了目前对它们的实时分析和处理面临着巨大的挑战.这就需要开展高通量图像视频计算方面的研究,需要结合新型硬件结构,利用其体系结构优势,提出一系列实用的高通量图像视频计算理论与方法,提升数据中心的图像视频数据处理效率.为此,在详细地分析了现有的高通量图像视频计算相关方法与技术的基础上,探讨了现有高通量图像视频计算方法研究的不足;进一步地,分析了高通量图像视频计算的3个未来研究方向:高通量图像视频计算理论、高通量图像视频分析方法及高通量视频编码方法.最后,总结了高通量图像视频计算需要解决的3个关键科学问题.这些问题的解决将为互联网图像视频内容监管、大规模视频监控、图像视频搜索等重要应用提供关键技术支持.
[Abstract]:Image and video data on the Internet are being generated and spread at full speed. These data not only have large scale, but also have the remarkable characteristics of high concurrency, high dimension and large traffic, which lead to the great challenge to their real-time analysis and processing. Therefore, it is necessary to carry out the research on high-throughput image and video computing, and combine the new hardware structure with the advantage of its architecture, and put forward a series of practical theories and methods of high-throughput image and video computing. Improve the efficiency of image and video data processing in data center. Therefore, based on the detailed analysis of the existing high-throughput image and video computing methods and technologies, this paper discusses the shortcomings of the existing high-throughput image and video computing methods. Furthermore, three future research directions of high-throughput image and video computing are analyzed: high-throughput image video computing theory, high-throughput image video analysis method and high-throughput video coding method. Finally, three key scientific problems to be solved in high throughput image and video computing are summarized. The solution of these problems will provide the key technology support for the Internet image video content supervision, large-scale video surveillance, image video search and other important applications.
【作者单位】： 南京理工大学计算机科学与工程学院;中国科学院计算技术研究所;
【基金】：国家“九七三”重点基础研究发展计划基金项目(2014CB347600) 国家自然科学基金项目(61402228);国家自然科学基金优秀青年科学基金项目(61522203)~~
【分类号】：TN919.81;TP391.41
[Abstract]:Image and video data on the Internet are being generated and spread at full speed. These data not only have large scale, but also have the remarkable characteristics of high concurrency, high dimension and large traffic, which lead to the great challenge to their real-time analysis and processing. Therefore, it is necessary to carry out the research on high-throughput image and video computing, and combine the new hardware structure with the advantage of its architecture, and put forward a series of practical theories and methods of high-throughput image and video computing. Improve the efficiency of image and video data processing in data center. Therefore, based on the detailed analysis of the existing high-throughput image and video computing methods and technologies, this paper discusses the shortcomings of the existing high-throughput image and video computing methods. Furthermore, three future research directions of high-throughput image and video computing are analyzed: high-throughput image video computing theory, high-throughput image video analysis method and high-throughput video coding method. Finally, three key scientific problems to be solved in high throughput image and video computing are summarized. The solution of these problems will provide the key technology support for the Internet image video content supervision, large-scale video surveillance, image video search and other important applications.
【作者单位】： 南京理工大学计算机科学与工程学院;中国科学院计算技术研究所;
【基金】：国家“九七三”重点基础研究发展计划基金项目(2014CB347600) 国家自然科学基金项目(61402228);国家自然科学基金优秀青年科学基金项目(61522203)~~
【分类号】：TN919.81;TP391.41
【相似文献】
相关期刊论文 前10条
1 陆l，
本文编号：2311208
本文链接：https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/2311208.html
