图像数据库分类的自组织技术

本文采用 Kohoenmap自组织神经网络技术 ,提出一种依据检索特征对图像数据库进行预分类的方法 .在文中利用这种预分类实现按类别浏览图像数据库 (image browsing by category)的功能 ,提出一种层次化的类 SS- tree的二分索引结构 .实现建立在预分类基础上的示例查询 .这种查询方法与穷搜索方法相比 ,具有更高的检索效率 ,检索次数随着库规模的增加不会显著增多 .基于图像库的实验验证了分类算法和检索算法的有效性

Self-Organizing Technique for Categorizing Image Database

In this paper, a cluster technique based on Kohonenmap neuron network is used to categorize the images in database. And image database browsing technique is realized based on image category. Also an index structure similar as SS-tree is proposed. Query by example is done on the image category, it is more efficient than exhaustive search method, and the compare time don't increase apparently with database size increasing. The experiments demonstrate the efficiency of our cluster algorithm and the proposed query by example method.