摘要
传统的空间金字塔匹配方法时间复杂度较高,其所采用的SIFT底层特征缺少颜色信息,从而导致图像分类性能不佳。该文提出了一种融合颜色和尺度不变特征的CSIFT算子,通过建立CSIFT词典的有向图邻接矩阵,对词典中单词的距离进行度量,构建了n阶距离度量矩阵,对图像进行相似性度量并分类。实验结果表明,该方法在优化图像词典构造方面有明显效果,提高了图像分类精度。
The traditional SPM(Spatial pyramid matching)method suffers from the high computation problem,and meanwhile the SIFT adopted by SPM lost sufficient color information which leads to the poor performance on image classification.In this paper,a novel CSIFT descriptor based on color and SIFT is proposed,which construct directed graph adjacency matrix to measure the distance of visual words,and construct n-order distant matrix to measure the image by similarity and classify the images.The experimental results show that this method has obvious effect in optimizing the construction of image dictionary and improves the accuracy of image classification.
作者
李青彦
彭进业
李展
LI Qingyan;PENG Jinye;LI Zhan(School of Electronics and Information,Northwestern Ploytechnical University,Xi′an 710072,China;School of Information Science and Technology,Northwest University,Xi′an 710127,China)
出处
《西北大学学报(自然科学版)》
CAS
CSCD
北大核心
2018年第1期50-56,共7页
Journal of Northwest University(Natural Science Edition)
基金
中国博士后科学基金资助项目(2014M552478)