摘要
基于内容的图像检索技术可以从庞大的网络图像库中较快速地找到用户所要查询的图像,但单一视觉特征不能充分地刻画图像内容信息等问题降低了检索结果的精度.针对于此,该文提出了一种基于Zernike分布矩与Contourlet变换相融合的彩色图像检索算法,首先算法对彩色图像进行Zernike分布矩提取,然后进行Contourlet变换,计算图像转化后的方差和熵,将结果作为图像的纹理特征,归一化处理并计算彩色图像纹理特征的权值.最后将计算后的纹理特征作为图像检索的依据,以此进行彩色图像的检索.仿真实验证明,这种基于Zernike分布矩与Contourlet变换相融合的彩色图像检索算法能够较全面地描述图像的语义信息,更具层次地依据彩色图像的本质特征进行彩色图像检索,一定程度上提高了图像检索的查准率和查全率.
Using the content-based image retrieval technology(Content-Based Image Retrieval,CBIR)can be more quickly find the user to query image from the vast network of image library,but a single visual features can not adequately describe the image content information and other issues,thus reducing the accuracy of the search results.In light of this,this paper presents a fusion of a variety of image retrieval algorithm,the method first construct a chromaticity distribution Zernike moments for color image feature extraction,and then use the Contourlet transform multi-directional image multiscale decomposition,and calculating the variance and entropy of each sub-band decomposition,as texture features of images,then these features and calculating normalized weights corresponding to each feature.Finally,calculate the similarity between the images of these features,the search result is returned and sorted.Simulation results show that this multi-feature fusion method can multi-level search algorithm describes the semantic information of the image,to some extent,improve the image retrieval precision and recall.
出处
《华中师范大学学报(自然科学版)》
CAS
北大核心
2015年第2期190-194,共5页
Journal of Central China Normal University:Natural Sciences
基金
河南省科技攻关项目(122102210549
132102210423)