一种快速分类搜索运动估计新方法

在视频编码中 ,基于块的运动估计算法被广泛应用 .在保证估计质量的前提下 ,为了降低运动估计算法的搜索次数 ,提出了一种对于不同类型的块采用不同的搜索范围和搜索步骤的分类快速搜索 (CFS)运动估计新算法 .该算法首先对块进行分类 ,然后确定其搜索范围和搜索步骤 ,在应用分类搜索法时 ,根据运动矢量的中心偏置特性 ,将第 1步和第 2步的搜索窗采用 5× 5的窗口 ,第 3步采用 3× 3的窗口 .结果表明 ,该分类快速搜索新算法在运动矢量的估计质量上 ,明显优于传统三步搜索法 ,且搜索次数与传统三步搜索法相比 ,降低了 2 3% ,与全搜索法相比 ,降低了 91% .实验结果证明 ,该算法尤其适用于快速运动、复杂运动序列的运动估计 .与传统的全搜索法和三步搜索法相比 ,其更适合于用硬件实现 .

A Fast Classified Search Algorithm for Motion Estimation

For video coding, the algorithm of motion estimation based on blocks is widely used. Under the prerequisite of assuring estimation accuracy, a fast classified search method applying different search range and different search steps for different blocks is given in this paper. According to the criterion of minimization of mean square error(MSE) between two frames, two threshold valuesT1andT2are determined. Each block's MSE is compared withT1andT2. Then blocks' movement is classified into fast, slow and still moving. Different search range and different search steps are applied for different blocks. For fast moving blocks three steps search is used. For slow moving blocks two steps search is used. For still blocks one step search is used. Based on the center-biased property of the motion vectors, the first and second search step use the 5×5 window and the third search step uses a 3×3 window. The results show that this method is better in quality than 3SS and reduce 3SS search steps by 23% and FS search steps by 91%. Furthermore this algorithm is more suitable for VLSI implement as well.