论文部分内容阅读
网格模型的运动估计和跟踪在基于模型的编码方法中占有重要的地位 ,在现有的网格运动估计技术中 ,六边形匹配 (HexagonalMatching)是一种有效的网格模型运动估计算法 ,它采用局部最优和迭代策略相结合的方法得到较好的全局运动估计结果 .六边形匹配算法由块搜索和迭代六边形匹配两步组成 ,同所有的基于模型的运动估计技术一样 ,六边形匹配算法也存在运算量大的问题难于满足实时编码需要 .本文对六边形匹配算法从三个方面进行改进 :用四步搜索替代全搜索、快速的纹理映射技术和有效的节点排序减少迭代次数 .改进后的算法在基本上保证原算法的运动估计效率的情况下 ,算法的复杂度有明显的降低 ,使得基于模型的编码技术向实用化前进了一步 .
Mesh model motion estimation and tracking plays an important role in the model-based coding methods. In the existing grid motion estimation techniques, Hexagonal Matching is an effective mesh model motion estimation algorithm, It uses a combination of local optimal and iterative strategies to get good global motion estimation results.Hygon matching algorithm consists of block search and iterative hexagonal matching two steps, as with all model-based motion estimation techniques, Hexagonal matching algorithm also has the problem of large amount of computation, which is difficult to meet the needs of real-time coding.This paper improves the hexagonal matching algorithm from three aspects: four-step search instead of full search, fast texture mapping and effective node sorting Reducing the number of iterations.The improved algorithm basically guarantees the efficiency of the original algorithm in the case of the motion estimation, the complexity of the algorithm has significantly reduced, making the model-based coding technology to practical one step forward.