Motion estimation (ME) process is the most computational complex part of a video encoder. It exploits temporal redundancy present in a video sequence to achieve compression. To make the video encoding process in real time, alleviating the computational burden of motion estimation is mandatory. In this work, an efficient hierarchical diamond search (EHDS) ME algorithm is being presented. This algorithm consists of multiresolution hierarchies, modified diamond search and modified early jump-out mechanism. Multiresolution hierarchies and the modified diamond search algorithm help either locate the global minima or some location nearby it, as best match by first finding starting point of search near the global minima which improves quality and reduces complexity, while the modified early jump-out mechanism further reduces complexity. Results have shown that the proposed algorithm has less complexity and power consumption than the conventional diamond search (DS) and other existing DS algorithms, while the quality is comparable to full search (FS) ME algorithm.

运动估计 (ME) 过程是视频编码器中计算最复杂的部分。它利用视频序列中存在的时间冗余来实现压缩。为了使视频编码过程实时化，必须减轻运动估计的计算负担。在这项工作中，提出了一种有效的分层菱形搜索 (EHDS) ME 算法。该算法由多分辨率层次结构、改进的菱形搜索和改进的早期跳出机制组成。多分辨率层次结构和改进的菱形搜索算法有助于定位全局最小值或其附近的某个位置，通过首先在全局最小值附近找到搜索起点来进行最佳匹配，从而提高质量并降低复杂性，而改进的早期跳出机制进一步降低复杂性。