中文摘要：

提出了一种2 bit深度像素的运动估计算法。首先,将像素深度的降采样过程形式化为区间分划和区间映射2个步骤,其中前者为多对一映射,决定着运动估计性能,后者为一一映射;其次,提出一种非均匀量化方法求解区间分划的3个初始阈值,并利用隶属度函数对初始阈值细化,从而克服信号噪声等因素导致的初始阈值周围像素值的误匹配;再次,讨论了适用于2 bit深度像素运动估计的误差度量准则,进而提出了基于模糊量化和2 bit深度像素的运动估计算法;最后,借助信号自相关函数,建立比特深度转换误差—运动向量精度模型来估计该算法所能达到的预测精度。实验结果证明,对于多种类型的视频序列,尤其是场景细节和物体运动比较复杂者,该算法始终能保持较高的估计精度,运动补偿的平均峰值信噪比较之传统2 bit深度像素的运动估计提高0.27 dB。

英文摘要：

A motion estimation algorithm was proposed using 2 bit-depth pixels. The reduction of pixel depth was first formalized by two successive steps, namely interval partitioning and interval mapping. The former is a many-to-one mapping which determines motion estimation performance, while the latter is a one-to-one mapping. A non-uniform quantization method was then presented to compute three initial thresholds of the interval partitioning. These initial thresholds were subsequently refined by using a membership function to solve the mismatch of pixel values near them caused by signal noise and so on. Afterwards, a matching criterion was discussed suitable for the motion estimation using 2 bit- depth pixels. A novel motion estimation algorithm was consequently addressed based on 2 bit-depth pixels and fuzzy quantization. To further predict the precision of the proposed algorithm, a bit resolution reduction error-motion vector precision model was built by exploiting the auto-correlation function. Extensive experimental results show that the pro- posed algorithm can always achieve high motion estimation precision for video sequences with various characteristics, especially for those with detailed scene and complex motion. Compared with traditional 2 bit motion estimation, the pro- posed algorithm gains 0.27 dB improvement in terms of average peak signal-to-noise ratio of motion compensation.