基于重叠亮度序的曲线描述子  

Overlapping intensity order-based curve descriptor

作　　者：江燕 王静[1] 刘红敏[1] JIANG Yan;WANG Jing;LIU Hongmin(College of Computer Science and Technology,Henan Polytechnic University,Jiaozuo 454000,Henan,China)

机构地区：[1]河南理工大学计算机科学与技术学院,河南焦作454000

出　　处：《河南理工大学学报（自然科学版）》2020年第3期114-121,共8页Journal of Henan Polytechnic University（Natural Science）

基　　金：河南省科技攻关项目(182102210053)。

摘　　要：针对现有亮度序划分方法导致子区域划分不稳定的问题,在已有的亮度序均值标准差描述子(IOMSD)和亮度序曲线描述子(IOCD)基础上,引入重叠亮度序划分思想,提出基于重叠亮度序的曲线描述子OIOMSD和OIOCD。与传统描述子构造子区域划分互不重叠不同,本文首先依据图像局部区域内各像素点的亮度序划分原始子区域,其次将每个子区域与其后相邻的子区域进行一定程度的重叠合并,成为最终的子区域。该方法不仅克服了固定形状划分子区域产生的边界误差,而且能够解决噪声和非单调光照对子区域边界点的影响。实验结果表明,OIOMSD和OIOCD对旋转变化、视角变化、噪声和光照变化图像具有鲁棒性,尤其是OIOCD在非单调光照图像上的匹配性能优于已有的IOCD算法。To solve the problem that instability of sub-region division caused by existing intensity order methods,based on intensity order mean-standard deviation descriptor(IOMSD)and intensity order curve descriptor(IOCD),two curve descriptors OIOMSD and OIOCD were proposed by introducing the idea of overlapping intensity order.Different from traditional descriptors whose sub-regions division was not overlapped,original sub-regions were firstly partitioned according to the intensity order of all pixel points in the image local region,then each sub-region and the next one are partial overlapped to form final sub-regions.This method not only overcame the boundary error caused by the fixed shape sub-regions,but also could solve the influence of noise and non-monotonic illumination on the sub-region boundary points.Experimental results showed that OIOMSD and OIOCD had robust performance under image rotation,viewpoint change,noise and illumination change.Moreover,the matching performance of OIOCD under non-monotonic illumination images was better than IOCD.

关 键 词：亮度序均值标准差描述子 亮度序曲线描述子 曲线匹配 重叠亮度序 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]