基于对角线偏移的SURF算子改进  

An Improvement of SURF Based on Diagonal Offset

作　　者：谭松[1] 高珏[1] 刘有科[1] 许华虎[1] 

机构地区：[1]上海大学,上海200444

出　　处：《计算机技术与发展》2014年第11期1-4,共4页Computer Technology and Development

基　　金：国家重大科技专项课题(2009ZX04001-111)

摘　　要：由于SURF算子对光照仅有部分的不变性,从而在光照条件变化大的环境下可能会导致提取的特征数与匹配数减少的不足。为了克服该不足,文中深入分析光照变化对SURF算子对特征提取的影响,并提出利用对角线偏移模型对SURF算子进行改进,统一输入图像的光照信息,使得改进后的特征算子对光照的变化具有更强的不变性,从而使得改进后的算法在较大光照变化的条件下依旧能够稳定地提取特征点及匹配。Since SURF only is partly invariant for illumination,if it is applied into the variable illumination changing scene,the perform-ance of features and matching will decrease highly. In order to overcome the drawback,make a deeper analysis about how the illumination influences the performance of SURF,and improve the original SURF with uniform illumination by combining the diagonal-offset model. The improved algorithm has a stronger ability about invariant for illumination,so that can get stable feature points even under a big illumi-nation changing and match.

关 键 词：加速鲁棒特征 特征提取 光照 对角线偏移 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]