首页 | 本学科首页   官方微博 | 高级检索  
     

MPEG-4码率控制方案的改进研究
引用本文:张忠伟,刘贵忠. MPEG-4码率控制方案的改进研究[J]. 西安交通大学学报, 2005, 39(10): 1052-1055,1063
作者姓名:张忠伟  刘贵忠
作者单位:西安交通大学电子与信息工程学院,710049,西安
基金项目:国家自然科学基金资助项目(60272072),教育部高等学校博士学科点专项科研基金资助项目(2000069828)
摘    要:针对低码率、低时延视频通信中MPEG-4 Q2码率控制方案导致的图像质量波动问题,利用拉格朗日乘子法,提出了一个率失真优化的目标比特估计算法(LAG-BA),根据前一帧图像的编码信息和虚拟缓存器状态预测当前帧的目标比特数.LAG-BA不但降低了连续图像质量变化的幅度,而且保证了算法与输入图像内容的自适应性.针对LAG-BA的误差累积效应,根据最小二乘支持向量机的原理,进一步提出了一个考虑到多个相关图像编码信息的目标比特估计算法(LSSVM-BA),并用LAG-BA和LSSVM-BA取代MPEG-4 Q2方案中的比特分配算法设计了两个改进的MPEG-4 Q2方案.实验结果表明,与原始的MPEG-4 Q2方案相比,两种改进的MPEG-4 Q2方案不但取得了更加平稳的图像质量,而且分别使编码的峰值信噪比提高了0.12 dB和0.44 dB.

关 键 词:目标比特估计  拉格朗日乘子  最小二乘支持向量机
文章编号:0253-987X(2005)10-1052-04
收稿时间:2004-11-25
修稿时间:2004-11-25

Improvements for MPEG-4 Rate Control Scheme
Zhang Zhongwei,Liu Guizhong. Improvements for MPEG-4 Rate Control Scheme[J]. Journal of Xi'an Jiaotong University, 2005, 39(10): 1052-1055,1063
Authors:Zhang Zhongwei  Liu Guizhong
Abstract:In order to decrease the picture quality fluctuation due to Q2 rate-control scheme in low rate and low delay video communication,firstly,a target bit-allocation algorithm was proposed with Lagrange multiplier method,where the coding information of previous one frame and the buffer status were taken into account.The proposed algorithm is denoted Lagrange bit allocation(LAG-BA),which not only reduces the quality fluctuation between continuous frames,but also keeps its adaptability for input images.Secondly,to reduce the error cumulating effect of LAG-BA,another bit-allocation algorithm was derived from the least square support vector machine,where the coding information of several similar frames was taken into account.The proposed algorithm is denoted least square support vector machine bit allocation(LSSVM-BA),which further depresses the video quality fluctuation.Replacing bit allocation algorithm in MPEG-4 Q2 scheme with LAG-BA and LSSVM-BA, two improved MPEG-4 Q2 schemes were designed.Experimental results show that the improved Q2 schemes not only produce more stable image quality,but also increase peer-signal-noise-rate by 0.12 dB and 0.44 dB,respectively,comparing with the original Q2 scheme.
Keywords:target bit-estimation  Lagrange multiplier  least square support vector machine
本文献已被 CNKI 维普 万方数据 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号