Abstract: As the existing methods for video mosaic take high computational costs，a fast and effective algorithm based on SURF feature for video mosaic is proposed.The algorithm uses SURF method with strong robustness and superior performance to extract feature instead of SIFT.At the aspect of feature matching，a novel matching scheme based on hash mapping and bidirectional nearest neighbor distance ratio is presented，which can quickly and effectively obtain the relationship between the features.In order to exclude the error matchings，a RANSAC technique is applied to eliminate outliers to ensure effectiveness of the matched pairs，and then the global motion parameters are estimated by a least-squares solution，finally panorama from video sequence is achieved using the parameters.Experimental results show that the method with strong robustness performs fast and effectively and has highly valuable in practice.

Key words: video mosaic, Speeded Up Robust Feature（SURF）, feature matching, motion parameters estimation, image registration

摘要： 针对现有的视频图像序列拼接方法处理速度慢的问题，提出一种基于SURF特征的快速有效的拼接算法。该算法用鲁棒性强且计算性能优越的SURF算子取代传统的SIFT算子进行特征点提取；在特征点匹配方面，提出了一种基于哈希映射和双向最近邻距离比的匹配算法，可以快速有效地获得特征点间的对应关系。为了消除由于运动物体干扰带来的误匹配，采用随机采样一致性（RANSAC）方法来消除外点确保匹配的有效性，再通过最小二乘法估计视频帧之间的全局运动参数，最终拼接形成全景图。实验结果表明，该拼接算法快速有效，鲁棒性强，具有较高的使用价值。

关键词: 视频拼接, SURF, 特征点匹配, 运动参数估计, 图像配准