特征融合和支持向量机反馈的图像检索算法  

IMAGE RETRIEVAL ALGORITHM BY INTEGRATING FEATURES FUSION AND SUPPORT VECTOR MACHINE FEEDBACK

作　　者：罗益荣[1] 

机构地区：[1]湖南农业大学信息科学技术学院,湖南长沙410126

出　　处：《计算机应用与软件》2014年第10期209-212,共4页Computer Applications and Software

基　　金：湖南省科技计划项目(2011FJ3055)

摘　　要：检索算法是海量图像自动检索基础。鉴于单一特征无法准确描述图像内容,结合时域和频域纹理特征优点,提出一种特征融合和支持向量机反馈的图像检索算法。首先取图像的LBP直方图作为空域特征,并利用Brushlet变换提取子带能量特征作为频域特征;然后采用马氏距离相似度量进行图像初步检察;最后采用支持向量机反馈提高图像检索准确率。仿真结果表明,相对于单一特征检索算法,该图像检索算法提高了图像检索的平均准确率,可以更准确地查找到用户所需的图像。Retrieval algorithm is the basis of the massive image automatic retrieval. Since the single feature cannot accurately describe the image content, we propose an image retrieval algorithm by integrating the feature fusion and support vector machine feedback which combines the time domain features with frequency domain features. Firstly, the LBP histogram of the image is taken as the spatial features while the Brushlet transform is employed to extract the sub-band energy features to be the frequency domain features, then the Mahalanobis distance similarity measurement is used to make initial detection on the images, and finally the support vector machine feedback is adopted to improve the accuracy rate of image retrieval. Simulation results show that relative to the single feature retrieval algorithms, the proposed algorithm improves the average accuracy rate of the image retrieval, and can find the user desired image more accurately.

关 键 词：图像检索 纹理特征 局部二值模式 BRUSHLET变换 支持向量机反馈 

分 类 号：TP85[自动化与计算机技术—检测技术与自动化装置]