基于颜色不变量和仿射不变性的彩色图像配准  

Registration of color images based on color invariants and affine invariant

作　　者：刘小丹[1] 吴笑嫦 曹耐[1] 孙红岩[1] 

机构地区：[1]辽宁师范大学计算机与信息技术学院,辽宁大连116081

出　　处：《计算机工程与设计》2014年第12期4243-4248,共6页Computer Engineering and Design

基　　金：国家自然科学基金项目(41271422);辽宁省教育厅自然科学基金项目(L2012379)

摘　　要：ASIFT图像配准算法实现了完全的仿射不变性,但色彩信息的缺失导致算法对光照条件敏感。结合颜色不变量理论和ASIFT算法,提出一种彩色图像配准算法CASIFT。在ASIFT基础上以颜色不变量代替灰度信息提取特征点并建立特征点的特征描述子,解决了ASIFT算法在光照差异较大情况下彩色图像配准率低的问题。实验结果表明,CASIFT彩色图像配准算法对于光照变化和视角变化均有更好的鲁棒性。The ASIFT image registration algorithm achieves completely affine invariant,but the lack of color information leads to the sensitivity to lighting conditions.The color invariant theory was applied to the ASIFT algorithm,a new color image registration algorithm CASIFT was proposed.Instead of using the gray-scale information,color invariants were used to extract feature points,the feature descriptor of feature points was established based on color invariants,the problem that the ASIFT algorithm fails to match images under serious different lighting conditions was solved.Experimental results show that,combining the color invariants,the proposed algorithm is more robust in different lighting conditions and viewing angles.

关 键 词：图像配准 尺度不变特征变换 颜色不变量 仿射不变性 仿射不变的SIFT 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]