基于两种纹理特征聚类的图像检索

A Texture Cluster Indexing Scheme for Content-based Image Retrieval

设计和实现两种不同的分形维数作为纹理特征进行聚类的方法。提取几百幅不同图像的两种纹理特征,对特征库按聚类算法建立索引结构,形成图像的分类库,通过两种不同纹理特征的检索与无聚类的特征检索相比,实验结果表明聚类方法大大缩短了检索时间。结论:作为海量图像的检索,有效的图像特征结合聚类是一个有力工具,在研究信息分类与识别等方面具有应用潜力。

A new efficient cluster algorithms of texture feature of two different fractal dimension is proposed , Methods extraction large- scaled image two texture features, after these features were extracted, feature database is indexed by clustering . The results of a series of tests show there is correct about effective and efficient between the without clustering for image retrieval and with clustering for it. Conclusion As a powerful tool for large image to use clustering and it has potential on study classification and information recognition.