• 中国计算机学会会刊
  • 中国科技核心期刊
  • 中文核心期刊

计算机工程与科学

• 论文 • 上一篇    下一篇

基于融合特征的SURF配准算法

秦煜1,吴静静1,2,安伟1,2,张美娟3   

  1. (1.江南大学机械工程学院,江苏 无锡 214122;2.江南大学江苏省食品先进制造装备技术重点实验室,江苏 无锡 214122;
    3.无锡职业技术学院,江苏 无锡 214122)
  • 收稿日期:2016-04-07 修回日期:2016-08-15 出版日期:2017-10-25 发布日期:2017-10-25
  • 基金资助:

    国家自然科学基金(61305016)

A SURF image registration algorithm based on fusion features

QIN Yu1,WU Jing-jing 1,2,AN Wei1,2,ZHANG Mei-juan3   

  1. (1.School of Mechanical Engineering,Jiangnan University,Wuxi 214122;
    2.Jiangsu Key Laboratory of Advanced Food Manufacturing Equipment and Technology,Jiangnan University,Wuxi
    214122;
    3.Wuxi Institute of Technology,Wuxi 214122,China)
  • Received:2016-04-07 Revised:2016-08-15 Online:2017-10-25 Published:2017-10-25

摘要:

SURF算法是一种尺度不变、旋转不变且鲁棒性良好的配准算法,但其丢失了图像的颜色特征,因此对于彩色图像的配准效果
不佳。为此,提出了一种基于融合特征的SURF配准算法。该算法首先利用彩色图像的颜色不变量和DLBP纹理特征构造融合特
征灰度图,并提出了一种基于彩色图像颜色直方图的自适应方法来调节融合特征的权重;然后,利用SURF算法在融合特征灰
度图上进行特征点的提取与匹配;最后,使用改进的RANSAC算法去除误匹配点。实验结果表明,对于彩色图像,此算法有效
地增加了提取的特征点数,并加快了配准速率。

关键词: SURF算法, 颜色不变量, DLBP纹理特征, RANSAC算法

Abstract:

The speed-up robust features (SURF) algorithm is a scale-invariant, rotation-invariant and robust
registration algorithm. However, since the color features of the image can be lost, its registration effect
of color images is poor. We therefore propose a SURF registration algorithm based on fusion features.
Firstly, we utilize the color invariant and the double local binary pattern (DLBP) texture features of the
color image to construct the grayscale of fusion features. Meanwhile, we propose an adaptive method to adjust
the weights of fusion features based on the color histogram of the color image. Then, the SURF algorithm is
used to extract and match the feature points on the grayscale of fusion features. Finally, the improved
random sample consensus (RANSAC) algorithm is used to remove the mismatch points. Experimental results show
that the proposed algorithm can effectively increase the feature points and speed up the registration rate
for color images.

Key words: SURF algorithm, color invariant, DLBP texture feature, RANSAC algorithm