基于深度学习的视频多目标码率分配策略
DOI:
作者:
作者单位:

作者简介:

通讯作者:

中图分类号:

TP391

基金项目:


Deep learning based multi-target on video rate allocation strategy
Author:
Affiliation:

Fund Project:

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    在高效视频编码(HEVC)中,最新的λ域码率控制方案自适应地对每个像素分配比特数,在自适应分配比特数的过程中,不考虑目标对象的编码质量。针对该问题提出了基于深度学习的视频多目标码率分配方案。在该方案中,首先采用目标检测算法SSD来检测识别编码视频中的多目标对象,获取对象信息。然后在帧级码率分配过程中,根据帧间差异,采用自适应比特比率,为每帧视频分配合理的比特数。最后,以更小的量化参数(QP)来提升目标对象编码质量。实验结果表明,与HM16.9相比,改进的算法在码控精度几乎不变的情况下,对全局的平均质量提升0.15 dB, 目标区域的质量提升0.35 dB。

    Abstract:

    In high efficiency video coding HEVC, the latest λdomain rate control scheme adaptively allocates the number of bits for each pixel. In the process of adaptively allocating the number of bits, the encoding quality of the target object is not considered.Aiming at this problem, a video multi target rate allocation scheme based on deep learning is proposed. In this scheme, a target detection algorithm single shot multibox detector (SSD) is first used to detect and identify multi objects in a coded video to obtain object information. Then in the frame level code rate allocation process, according to the difference between frames, an adaptive bit rate is used to allocate a reasonable number of bits for each frame of video. Finally, the quantization quality of the target object is improved with a smaller quantization parameter (QP). The experimental results show that compared with HM16.9, the proposed algorithm improves the global average quality by 0.15 dB and the quality of the target area by 0.35 dB when the code control accuracy is almost constant.

    参考文献
    相似文献
    引证文献
引用本文

朱丽莎,王国中,滕国伟,杨郑龙,张立亮.基于深度学习的视频多目标码率分配策略[J].电子测量技术,2019,42(2):96-102

复制
分享
文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:
  • 最后修改日期:
  • 录用日期:
  • 在线发布日期: 2021-07-08
  • 出版日期: