Motion estimation plays a very crucial role in many video applications. Hence, an accurate and fast algorithm is required in real-time applications. Although exhaustive block matching (EBM) algorithm is the best algorithm used for motion estimation in terms of performance, it is computationally very expensive. Several algorithms have been developed in the literature to reduce the computational time required by the EBM algorithm; however, their performance is not comparable to that of EBM. In this paper, we propose an extremely fast technique for motion estimation based on the edge map of only a fraction of the frame. Extensive experiments are carried out, and the results show that the proposed technique provides a performance comparable to that of the EBM algorithm, with a very low computational complexity.

运动估计在许多视频应用中起着至关重要的作用。因此，在实时应用中需要一种准确而快速的算法。尽管就性能而言，穷举块匹配（EBM）算法是用于运动估计的最佳算法，但它在计算上非常昂贵。文献中已经开发了几种算法来减少EBM算法所需的计算时间。但是，它们的性能无法与EBM媲美。在本文中，我们提出了一种仅基于一小部分帧的边缘图的极快运动估计技术。进行了广泛的实验，结果表明所提出的技术提供了与EBM算法相当的性能，并且计算复杂度很低。