基于方向局部极值模式的图像检索算法

发布时间：2018-05-31 05:23

本文选题：图像检索 + 局部二值模式　；参考：《计算机工程与设计》2016年12期

【摘要】：针对基于局部二值模式(local binary patterns,LBP)的图像检索算法对局部噪声敏感,导致算法抗噪能力弱、鲁棒性低等不足,提出一种基于方向局部极值模式(directional local extrema patterns,DLEP)的图像检索算法。通过计算图像像素点在0°、45°、90°和135°这个方向的局部极值,得到这4个方向的DLEP模式,利用DLEP在纹理提取过程中获取更多的方向边缘信息和空间关系;基于4个方向的DLEP模式,形成图像方向的DLEP直方图,根据DLEP直方图,构建图像的特征向量;引入相似距离度量进行特征匹配,完成图像检索。实验结果表明,与LBP、基于块的LBP算法和中心对称局部二值模式进行比较,在Corel图像数据库中,该算法具有更高的查准率、查全率和平均检索率。
[Abstract]:An image retrieval algorithm based on local binary pattern (LBP) is proposed, which is sensitive to local noise, which leads to weak anti-noise ability and low robustness. An image retrieval algorithm based on directional local extrema pattern (DLEP) is proposed. By calculating the local extremum of the pixel points in the direction of 0 掳~ 45 掳~ 90 掳and 135 掳, the DLEP mode in these four directions is obtained, and more direction edge information and spatial relationship are obtained by using DLEP in the process of texture extraction, and the DLEP mode based on four directions is used to obtain the spatial relationship. The DLEP histogram of the direction of the image is formed, the feature vector of the image is constructed according to the DLEP histogram, and the similarity distance metric is introduced to carry out the feature matching to complete the image retrieval. The experimental results show that compared with LBP, block based LBP algorithm and centrosymmetric local binary mode, the algorithm has higher recall, recall and average retrieval rate in Corel image database.
【作者单位】：南通理工学院软件工程系;南通大学杏林学院;南通大学计算机科学与技术学院;
【基金】：江苏省高校自然科学研究面上基金项目(16KJB520039) 南通理工学院科研课题基金项目(2014002)
【分类号】：TP391.41

【相似文献】