An adaptive motion estimation scheme using maximum mutual information criteria

Motion estimation in video coding can be formulated as an optimization problem. Recently, a motion estimation scheme that uses Renyi's error entropy as the optimization criterion, was proposed [1]. Motivated by [1], in this paper, we propose a different criterion in motion estimation, i.e., the criterion of maximum mutual information. Based on this new criterion, we design a motion estimation algorithm. Our results show that our algorithm achieves significantly lower computational complexity compared to existing fast-search methods for motion estimation. A salient feature of our algorithm is that it is ideally suited for wireless video sensor networks where limited bandwidth, restricted computational capability, and limited battery power supply pose stringent constraints on the system.

视频编码中的运动估计可表述为一个优化问题。最近，一种以雷尼误差熵作为优化准则的运动估计方案被提出[1]。受[1]的启发，在本文中，我们在运动估计中提出了一个不同的准则，即最大互信息准则。基于这个新准则，我们设计了一种运动估计算法。我们的结果表明，与现有的运动估计快速搜索方法相比，我们的算法计算复杂度显著降低。我们算法的一个显著特点是它非常适合无线视频传感器网络，在这种网络中，有限的带宽、受限的计算能力以及有限的电池供电对系统构成了严格的约束。