In this paper , we first established the image metadata used in our system which based on the famous dublin core , then we analyzed the abstraction and description visual features of image such as color texture and shape . next , we discussed the problem of similarity measure of visual feature , imported fuzzy logic into the distance feature and pointed out the disadvantages of geometry space based methods . for multi - dimension vector ' s high dimension nature , it ' s hard to index with traditional methods , we discussed how to lower the dimension using clustering and klt transformation 本文首先在dublincore的基礎上制定了適合我們要求的圖像元數據集;詳細分析了顏色、紋理、形狀等視覺特征的提取和表示方法;探討了圖像視覺特征相似度量的問題,將模糊技術引入直方圖的距離度量,分析了幾何空間距離度量函數的不足之處,提出了系統中采用的距離函數;針對圖像視覺特征向量的多維特性,分析了現有的各種降維技術和多維索引技術。