中文摘要：

文章提出了一种直接使用压缩域中的运动矢量进行全局运动估计的方法，并用遗传算法优化输入运动矢量与全局运动参数所产生运动矢量的平均匹配误差．最终的实验结果表明本文的方法能够很好的估计出全局运动的参数．并且提出了一种运用全局／局部运动信息进行视频中字幕遮挡区域的恢复的方法．实验结果表明该方法取得了较好的视觉效果．

英文摘要：

A Genetic Algorithm （GA） based Global Motion Eslimation （GME） method is proposed for compressed videos. Using motion vectors in compressed videos directly for GME, this approach utilizes GA to optimize the fitting errors between the input motion vectors and the ones generated from the estimated global motion parameters. Experimental results on different type of global motions show that the proposed genetic algorithm based global motion estimation method is satisfying. Moreover, a text-occluded region recovery approach is presented using the estimated global motion parameters,based on the Global/Local Motion（GM/ LM） information for text occluded region in the video sequences. Experimental results on different videos show that this method achieves better visual results.