基于分形维数的多线索商标图案检索系统  

Multi-clue Retrieval System for Trademark Images Based on the Fractal Dimension

作　　者：蒋金山[1] 李粤青[1] 汪国强[1] 

机构地区：[1]华南理工大学应用数学系,广东广州510640

出　　处：《华南理工大学学报（自然科学版）》2004年第4期37-42,共6页Journal of South China University of Technology(Natural Science Edition)

基　　金：广东省自然科学基金资助项目 (0 116 0 2 )

摘　　要：采用基于内容的图像检索技术 ,根据查询图像和目标图像的相似度匹配进行图像检索 .图像的内容由五种特征来描述 :图像的分形维数、分形矢量、边界分形维数、骨架分形维数和距离分布直方图 .系统界面友好、功能齐全 ,实验证明系统的检索效果比较理想 ,具有很好的平移、尺度和旋转不变性 ,并且只要查询图像在图库中 ,就一定能被检索出来 .另外 。The retrieval technique based on image content was adopted to retrieve images according to the similarity matching between queried images and target images. The image content is depicted by five features, namely, the fractal dimension and fractal vector of the image, the fractal dimensions of the image edge and the image skeleton, and the distance distribution histogram. This system is of the advantages of friendly interface and plenteous functions. Experimental results indicate that the system is of perfect retrieval effect on trademark images and is of excellent inflexibility of displacement, scale and rotation, and that the image can certainly be retrieved by the system if it exists in the image database. Besides, the system is proved to be robust.

关 键 词：分形维数 商标图案 图像检索 图像查询 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]