一种 HSV 空间上分层压缩感知的图像检索算法

摘要

By constructing a two-dimensional (2D)compressive sensing (CS)measurement model,a new image retrieval algorithm is proposed by extracting hierarchical HSV features and texture features. Firstly,the ideas of grid discrete partition and hierarchical mapping in HSV space are introduced,and hierarchical mapping matrix and similar GLCMin HSV grid space are defined.Secondly,the normalized Gauss random matrix is designed as measurement matrix,and compressive sampling on the above two ma-trixes is performed by 2D CS measurement model.With using PCA (Principal Component Analysis), the feature sequences as hierarchical HSV features and texture features are obtained from the above two hierarchical sampling matrixes.Finally,the above two features are infused to compute the overall similar-ity among images.Experimental results show that the above two features have good discrimination.It can improve the efficiency for image retrieval,and has good performance especially for images with complex backgrounds.%通过构建二维压缩感知测量模型，提出一种分层 HSV 特征和分层纹理特征提取与图像检索新算法。在图像 HSV 空间上引入网格离散划分和分层映射算子，定义一种基于 HSV 网格空间上的分层映射矩阵和拟灰度共生矩阵；设计归一化 Gauss 随机矩阵作为测量矩阵，使用二维压缩感知测量模型对上述两种矩阵进行压缩采样；采用 PCA （Principal Component Analysis）方法获取上述两种分层采样矩阵的特征值序列，作为图像的分层HSV 特征与分层纹理特征。最后融合上述两类特征综合计算图像间的整体相似度并实现图像检索。仿真实验表明，上述两类特征具有很好的可区分性，有效提高了图像检索效率，特别对复杂背景的图像检索性能更优。