图像纹理特征及其在CBIR中的应用

Texture Feature and Its Application in CBIR

基于内容的多媒体信息检索是当前世界的研究热点 ,然而在图像内容表示及其相似性度量这两个关键问题上取得的进展还不能令人满意 .通过深入研究纹理的本质和人类视觉系统的特性 ,结合MPEG 7标准提出的图像内容描述子概念 ,提出一种新的谱方法来描述图像的纹理特征 ,并将其应用于图像内容的表示和检索 .在此基础上 ,讨论了图像的颜色描述子和相似性度量策略 。

Content based image retrieval (CBIR) has been an active research area, however, the achievements in image representation and similarity measurement are not satisfying. After going deep into the nature of image texture and human vision system, incorporating the concept of image content descriptor proposed by MPEG 7, we propose a new texture spectrum to describe the texture feature of images. Moreover, we discuss color descriptor, image similarity scheme, and the combined color and texture descriptor. Experiment results and evaluation are illustrated.