Computer Engineering and Applications ›› 2015, Vol. 51 ›› Issue (5): 167-171.

Previous Articles     Next Articles

Research on novel optimization SIFT algorithm based fast mosaic method

WEI Lisheng1,2, ZHOU Shengwen1   

  1. 1.School of Electrical Engineering, Anhui Polytechnic University, Wuhu, Anhui 241000, China
    2.School of Mechatronics Engineering and Automation, Shanghai University, Shanghai 200072, China
  • Online:2015-03-01 Published:2015-04-08



  1. 1.安徽工程大学 电气工程学院,安徽 芜湖 241000
    2.上海大学 机电工程与自动化学院,上海 200072

Abstract: The Scale Invariant Feature Transform (SIFT) algorithm with low matching rate and long time is widely used in image registration. In order to improve the matching rate and the real-time performance, an improved SIFT algorithm is proposed, which extracts the extreme feature of image in the overlapping areas and down-sample the image registration. In feature matching process, the extreme consistency constraints of matching points in overlap region are considered. In order to reduce the description of operator dimension and false matching, and improve the matching accuracy, the local unipolar value in differential scale space is used. On the basis, the image scaling constraints with 180×180 are presented. According to the resized image and original image, the relationship between two transformation matrixes is derived to realize the rapid and accurate image registration. At last, experimental results show the effectiveness and feasibility of the proposed method.

Key words: Scale Invariant Feature Transform(SIFT), decimation, image matching, image stitching

摘要: 针对尺度不变特征变换(Scale Invariant Feature Transform,SIFT)算法图像配准时间长、匹配率低等问题,提出了重合区域图像极值特征提取法以及图像降采样特征配准法。在特征匹配的过程中,重点考虑重叠区域的特征匹配点对极值一致性约束条件,并利用差分尺度空间的局部单极值,以减小冗余特征点,节约特征提取与匹配时间;在此基础上,以图像尺度大小(选择180×180)作为缩放约束,对图像进行同比例插值缩小,并根据缩放后图像与原始图像变换矩阵之间的关系,计算出原始图像变换矩阵,实现图像的快速、精确配准。利用实例验证了所提方法的有效性和可行性。

关键词: 尺度不变特征变换(SIFT), 降采样, 图像匹配, 图像拼接