从视频读取帧，提取Sift特征点，进行特征点的匹配，合成视频-read frame from a video,get the sift feature,and match,then make a new video

用于视频压缩中的运动估计，由锚定帧输出预测帧、预测误差及PSNR与运行时间值-For video compression motion estimation, the prediction frame, the prediction error and the PSNR and run time value output by the anchor frame