摘要
将谱图理论引入图像检索领域,将原图像按照灰度级进行划分,针对各个图像灰度分布的差异提出自适应划分方法,把图像的每个划分都看作一个子图,利用谱图理论计算各个子图之间的Normalized cut(Ncut),生成一个Ncut矩阵,在生成的Ncut矩阵中,计算每一个子图与其余子图的相似关系,提取原图像的特征向量,达到检索的目的。实验结果表明,该方法优于传统方法。
Spectral graph theory is introduced into the field of image retrieval.The image is divided to sub-images according to gray levels,and adaptive cut algorithm is proposed to solve the difference of gray levels among images.Each cut is viewed as a sub-image,and the Normalized cut can be calculated among the sub-images via spectral graph theory and a Ncut matrix is made.The similarity is calculated between one sub-image and others and the eigenvector of the image is got using the Ncut matrix,achieving the purpose of image retrieval.
出处
《计算机工程与应用》
CSCD
北大核心
2011年第22期195-197,223,共4页
Computer Engineering and Applications
关键词
基于内容的图像检索
谱图理论
归一化划分
自适应
纹理
content-based image retrieval
spectral graph theory
Normalized cut
adaptive
texture
