基于深度学习与多线程资源分配的弹性视频编码方法研究
Research on Elastic Video Coding Method Based on Deep Learning and Multi-threaded Resource Allocation
针对复杂视频场景中动态内容对固定资源分配编码策略的挑战,提出一种基于深度学习与弹性资源调度的自适应视频编码方法。构建了轻量级时空特征提取网络(TS-Net),通过分析场景复杂度、运动矢量和帧间相关性,动态预测各编码单元的资源需求权重;在此基础上,设计了优先级驱动的线程池管理算法,依据TS-Net的输出实现计算资源的实时动态分配。实验结果表明,通过深度学习预测与弹性资源调度的协同优化,可使云转码带宽节省31.4%,移动端能效提升62.6%。
With the increasing complexity of video application scenarios, traditional fixed resource allocation encoding strategies are difficult to meet the needs of dynamic content. This study proposes an adaptive video coding method based on deep learning and elastic resource scheduling. A lightweight spatio-temporal feature extraction network (TS-Net) is constructed to dynamically predict the resource demand weights of individual coding units by analyzing scene complexity, motion vectors, and inter-frame correlation. Building on this, a priority-driven thread pool management algorithm is designed to achieve real-time dynamic allocation of computational resources based on the output of TS-Net. Experimental results demonstrate that the synergistic optimization of deep learning-based prediction and elastic resource scheduling significantly can reduce cloud transcoding bandwidth consumption by 31.4%and improve mobile energy efficiency by 62.6%.
| [1] |
安传华. 基于视频编码优化的中波转播传输系统设计[J]. 电视技术, 2025, 49(7): 142-144, 148. |
| [2] |
郭红伟, 朱策, 刘宇洋. 视频编码率失真优化技术研究综述[J]. 电子学报, 2020, 48(5): 1018-1029. |
| [3] |
罗文蔚, 高敏骐, 何慧华. 视觉与多模态方法论的框架与应用: 以视频内容分析为例[J]. 清华大学教育研究, 2025, 46(1): 136-147. |
| [4] |
于海, 杨磊, 高阳, |
| [5] |
刘美琴, 陈虹宇, 周一鸣, |
| [6] |
张旭光, 陈鸣锴, 魏昕. 算力网络支撑下的泛在化视频传输调度[J]. 计算机研究与发展, 2023, 60(4): 786-796. |
| [7] |
向海, 陈芬, 秦裔清, |
| [8] |
邹承益, 万帅, 朱志伟, |
| [9] |
兰尔铭,施隆照,宋佳柔, |
| [10] |
杨栩,郭红伟,李婉雪.基于熵平衡的全景视频编码[J].计算机应用研究,2023,40(6):1894-1899. |
2022年度福建省中青年教师教育科研项目(科技类)(JAT220713)
福建省教育科学“十四五”规划2022年度课题(FJJKGZ22-069)
/
| 〈 |
|
〉 |