摘要: 移动设备有限的电池能量是实时手语视频通信所面临的关键制约因素。为此,提出多优先级感兴趣区的H.264计算资源分配方法。从原始手语视频中快速检测出面部、双手和背景区域,依据视觉重要性的不同,给面部和双手区域分配较多的计算资源,给背景区域分配较少的计算资源。实验结果表明,该方法在保证手语视频感兴趣区编码质量的同时,可以大幅降低编码器的计算复杂度,减少电池能量消耗,延长能量受限设备的工作时间。
关键词:
人类视觉系统,
H.264标准,
感兴趣区,
计算资源分配,
手语视频,
电池能量
Abstract: The limitation of energy in battery powered devices is very challenging to apply H.264 to design a mobile sign language communication system. This paper therefore proposes a multiple priority Region of Interest(ROI) H.264 computation resource allocation scheme. In the system, the face, hands and background of sign language video is first detected with a fast algorithm. Then several coding parameters which allow the encoder to allocate computational resources differently are adaptively adjusted at the Macroblock(MB) level according to the relative priority of each MB. As a result, it keeps fine quality in face and hands, greatly reduces the computational complexity and prolongs the battery life.
Key words:
human visual system,
H.264 standard,
Region of Interest(ROI),
computation resource allocation,
sign language video,
battery power
中图分类号:
陈晓雷, 张爱华, 陈莉莉, 林冬梅, 杨欣翥. 多优先级感兴趣区H.264计算资源分配[J]. 计算机工程, 2013, 39(4): 283-286.
CHEN Xiao-Lei, ZHANG Ai-Hua, CHEN Chi-Chi, LIN Dong-Mei, YANG Xin-Zhu. H.264 Computation Resource Allocation in Multiple Priority Region of Interest[J]. Computer Engineering, 2013, 39(4): 283-286.