Abstract: Aiming at the problem that the general camera’s field of view is too small, by collecting two frame images from different cameras which have some overlap regions simultaneously, the SIFT algorithm is used to find the video frame image feature points; the Dichotomy Based Hash（DBH） algorithm is used to match the SIFT feature points; the Random Sample Consensus（RANSAC） algorithm is used to eliminate the false matches, and the mosaiced video can be obtained. Experiments show that this method can mosaic the video frame in real-time effectively. In addition, the method is feasible to low overlapped（even to 10%） video image.

Key words: Scale Invariant Feature Transform（SIFT） feature point, image registration, Dichotomy Based Hash（DBH）, real-
time, video mosaic

摘要： 针对普通摄像头水平视角较小的问题，通过同时采集具有一定重叠区域的摄像头视频帧图像，基于尺度不变特征变换（Scale Invariant Feature Transform，SIFT）特征点，用二分哈希搜索算法（Dichotomy Based Hash，DBH）进行匹配，用随机采样一致（Random Sample Consensus，RANSAC）算法消除误匹配，得到帧图像拼接映射关系。实验结果表明，该方法能有效地实现摄像头视频实时拼接，克服了既有方法在重叠区域小于20%时失效的不足，在重叠区域为10%左右时仍能取得有效的拼接。

关键词: 尺度不变特征变换（SIFT）特征点, 图像匹配, 二分哈希, 实时, 视频拼接