Computer Engineering and Applications ›› 2012, Vol. 48 ›› Issue (21): 244-248.

Previous Articles    

Application of SIFT algorithm in image registration of cigarette case

ZHU Meihua1, WEI Chengke2, ZHANG Suwei3   

  1. 1.Guangxi Zhuang Autonomous Region, Nanning 530022, China
    2.North China Institute of Computing Technology, Beijing 100083,China
    3.Taiji Computer Corporation Limited, Beijing 100083, China
  • Online:2012-07-21 Published:2014-05-19

SIFT算法在卷烟小包装图像配准中的应用

朱梅华1,魏承科2,张素伟3   

  1. 1.广西壮族自治区烟草专卖局,南宁 530022
    2.华北计算技术研究所,北京 100083
    3.太极计算机股份有限公司,北京 100083

Abstract: SIFT only considers local characteristics of the image during the searching of key points, therefore it is unable to extract representative feature points in processing image with complex texture background. To solve this problem, this algorithm considers the correlation between the feature points when extracting key points, reduces the dimension of feature description in the reference of SSIFT algorithm, and uses the statistical method to reduce the algorithm execution time. In this way, it can quickly extract the representative characteristics and filter out the key points of the texture pattern. The experiment results prove the efficiency and universality of the algorithm.

Key words: Scale-Invariant Feature Transform(SIFT), image registration, image matching, key point

摘要: 由于SIFT算法在寻找关键点时,只考虑了图像的局部特征,使得在具有复杂纹理背景的图像处理中,无法提取出具有代表性的特征点。针对这一问题,提出在提取关键点的时候,考虑特征点间的相关性,参照SSIFT算法缩小特征描述的维数,利用统计的方式缩短算法执行时间,使得算法能快速提取到具有代表性的关键点,滤掉纹理图案中的关键点。通过实验证明了算法的执行效率以及算法的普适性。

关键词: 尺度不变特征变换(SIFT), 图像配准, 图像匹配, 关键点