CBIR中图像综合特征向量的提取及其应用

A Extract and Measure Method of Image Multi-Feature Vector on CBIR

针对传统图像检索技术依赖图像单一视觉特征的不足,通过对图像单一特征向量的归一化处理获取图像综合特征向量,并采用加权欧氏距离进行图像相似性度量,再结合基于综合特征向量的相关反馈算法,可取得较理想的图像检索效果。

The traditional image retrieval methods are mainly dependent on the single vision feature,so the results are not always ideal.To cure this problem,the paper propose a new kind of image extract and measure method which is based on the multi-feature vector normalization and uses it as the standard of image＇s similarity measure.The method which is based on multi-feature vector and relevance feedback should have a very good effect.