图像检索中多个特征量的提取与比较

从不变量模型的原理出发，详细讨论颜色不变量、图像的纹理提取方法和彩色图像的梯度求解方法；应用这些方法，获取了12个特征量，并把它们应用于基于内容的图像检索系统中。通过大量实验和对不同特征量对应的检索结果的分析比较，得到了一些在基于内容的图像检索系统进行特征提取的结论：不同的检索内容需要相适应的不同的特征量；颜色量和颜色不变量的适用范围广，是很好的特征量；用从多次处理的图像中提取的特征量来检索效果较差；用多个特征量组合的特征进行检索能获得更好的检索效果。

Abstraction and Comparison of Multiple Features in CBIR System

In this paper,based on the principle of invariant model,we discuss in detail what the color invariantsare,and how to abstract the image texture and how to solute the gradient of multi-valued image,such as colorimage.And then,by using the method discussed above,we abstract 12 different features of an image.Everyfeature has its unique advantage and emphasizes a certain aspect of the image.As an application,we utilize thesefeatures in content-based image retrieval(CBIR) system to evaluate them.Finally,some con...