关键词: 图像纹理 图像检索 聚类空间模型 相似性

Abstract:

Texture is an important item of image information, texture-based image retrieval has been an active research area, and the similarity comparison of te xture features is a key to image retrieval. In this paper, based on the properties of texture, by expanding the Vector Space Model （VSM）, we create a   new model for matching the similarities of feature signatures to measure the similarities of texture images, which is called Clusters Space Model （CSM M）. And based upon it, a new method of multiple texture image retrieval without segmentation is shown. By comparing with other similarity measures for texture features, experimental results indicate it is consistent with human perception.

Key words: image texture knage retrieval CSM similarity