采用COL_(DIST)色差的一种图像检索方法研究

Study on An Image Retrieval Method Using COL_(DIST) Color Difference

在基于内容的图像检索方法中,颜色相似度的计算主要采用欧氏距离。然而,欧氏距离不符合人眼的色彩识别特征,导致检索准确率偏低。本文提出了一种采用COLDIST色差公式的组合特征检索方法。首先计算图像的颜色直方图作为颜色特征、灰度共生矩阵和灰度行程矩阵作为纹理特征、泽尼克矩作为形状特征。然后采用COLDIST色差公式计算颜色相似度,并结合纹理、形状特征的相似度计算图像相似度。利用图像数据库Corel10000对本文提出的方法进行仿真测试。结果表明该方法具有更好的检索性能,有效地提高了检索的准确率。

In image retrieval method based on content, the Euclidean distance is the most widely adopted to calculate the color similarity between images. However,the Euclidean distance is not consistent with the features of color recog-nition of human,and the retrieval accuracy is low. In this paper,a retrieval method based on combination features us-ing color difference formula （COLDIST） was proposed.Firstly, color histogram was extracted as color feature, GLCM （Gray Level Co-occurrence Matrix） and GLRLM （Gray Level Run-Length Matrix） were calculated as texture fea-tures, Zernike moment was calculated as shape feature. Then, COLDIST color difference was used to measure similarity of color features which was combined with similarities of texture and shape to get the image similarity. The proposed method was simulatively tested by utilizing Corel10000 database. The results showed that the proposed method was more efficient and had better retrieval performance.