为了提高块匹配运动估计快速算法的搜索速度、精度和鲁棒性,提出了一种多起点并发式快速运动估计搜索方法。根据匹配块搜索的特性,提出了均匀选取多个起始点的方法,同时根据搜索区域的大小制定了自适应的限制条件和进度表,使得算法的收敛速度有了很大的提高。该算法在迭代过程中不但能够接受使目标向好的方向前进的解,而且能够在一定限度内接受使目标恶化的解,这使得算法能够有效地跳出局部极小的陷阱。使用基于TMS320C6415芯片的DSP图像处理系统对抖动视频图像序列进行实验。实验结果表明,当视频图像为PAL制,分辨率为720×576个像素,并且相邻图像间的抖动范围为±40个像素时,使用新提出的方法,匹配参数的计算量不大于使用全域匹配方法计算量的1/8,并且搜索精度远高于其他传统的快速搜索方法。

A novel searching method of matching block based on the concurrent and multi-start algorithm is presentedto improve the search speed,accuracy and robustness in the fast operation of the block-matching motion estimation.Themethod of choosing starting points evenly is presented based on the characteristic of the block matching search,and the a-daptive conditions and adaptive schedule is established based on the searching region.In the processing of iteration,thenew method can not only adapt to the solution that lead the objective to the correct direction,but also adapt to the solutionthat have a little offset comparing with the objective.Therefore the concurrent and multi-start algorithm can be easy tokeep from the trap of local minima effectively.The image processing system based on the DSP chip of TMS320C6415isused to make the experiment for the video with shaking images.The results of experiment show that,when the televisionbroadcasting system of camera is PAL,the image resolution of video is 720×576pixels and the shaking motion betweenthe adjacent images is in the scope of±40pixels,the computational complexity of block matching parameters based onthe novel method is not more than one eighth of the computational complexity based on the full-search(FS)method,andthe searching accuracy of the novel method is much greater than the other traditional fast searching method.The methodcan improve the searching speed and achieve the high searching accuracy at the same time.