摘要
为了降低视频编码器的计算量,提出一种基于全零块的快速运动估计算法.该算法采用改进的逐次消除算法消除搜索过程中的无效匹配点,同时得到运动补偿预测残差块每个子块的直流系数,根据得到的直流系数和变换系数的分布模型推导出全零块的判断阈值.该算法不但能够加快运动估计的速度,而且还能降低离散余弦变换、量化、反量化和反离散余弦变换等的计算量,从整体上加快编码器的速度.实验结果表明:与已有快速运动估计算法相比,该算法编码时间最大可节约31%,而率失真性能基本无损失.
To speed up the video encoders further, a fast motion estimation algorithm based on the zero block is presented. This algorithm used the modified successive elimination algorithm to eliminate inviable candidate searching points and got DC transformed coefficients of each sub blocks in motion compensated prediction residual block, and then developed a criterion of zero block using the model of transformed coefficients and the gotten DC coefficients. It not only can speed up the motion estimation procedure, but also can save the computations of transform, quantization, inverse transform and inverse quantization. Simulation results show that: The encoder can obtain up to 31% reduction in computations versus some well-known fast algorithms with negligible rate distortion performance loss.
出处
《华中科技大学学报(自然科学版)》
EI
CAS
CSCD
北大核心
2008年第9期9-12,共4页
Journal of Huazhong University of Science and Technology(Natural Science Edition)
基金
湖北省自然科学基金资助项目(2005ABA255)
关键词
视频编码
运动估计
直流系数
变换系数模型
全零块
计算复杂度
video coding
motion estimation
DC (direct current) coefficients
transformed coefficients model
zero block
computation complexity