基于量子克隆选择的自适应多模式快速运动估计算法

该文提出一种新的基于量子克隆选择的自适应多模式快速运动估计算法。算法利用序列图像的时空预测运动矢量作为序列活动剧烈程度的依据，自适应选择搜索模式。静止块直接中止搜索；平缓运动类型块以(0，0)点为起始点，直接使用CDS搜索模式；剧烈运动类型块使用QCS/CDS联合搜索模式，首先使用量子克隆选择QCS寻找近似最优解，然后以近似最优解为起始点，使用CDS搜索。实验结果证明，算法在能够获得接近全搜索方法所得到的平均峰值信噪比前提下，平均搜索点数大大减少，搜索速度显著加快。

A Fast Multi-Pattern Motion Estimation Algorithm Based on Quantum Clonal Selection

A novel fast multi-mode motion estimation algorithm based on the quantum clonal selection is proposed. According to the predictive motion vectors derived from spatio-temporal coherence of video sequences, the type of searching modes can be confirmed adaptively. In the stationary macroblock mode, current search is terminated immediately. In the mild motion mode, CDS method is used directly. In the drastic motion mode, a combined QCS/CDS search strategy is utilized. Firstly, the QCS method is used to search the appropriate optimal motion vector, then the CDS method is used to search global optimal motion vector. Experimental results show that the algorithm can speed up the process of motion estimation with fewer search points, while maintaining the average PSNR by using the full search algorithm with little loss.