一种新的图像检索方法

常规基于内容图像检索的方法是提取图像的颜色、纹理等物理特征,运用相似性度量准则从图像库中查询相似的图像。为了提高图像检索的正确率,这里提出改进的方法。具体方法是:提取图像的物理特征,并将特征作为支持向量机(SVM)的输入向量,对图像进行分类,然后利用分类结果,对检索图像进行相似性匹配,从同类图像中找出相似的图像。实验结果显示,该方法的检索结果优于常规方法。

New Method for Image RetrieVal

The normal method for Content-based Image Retrieval(CBIR) is to extract color and texture features of images,then uses similarity measure to find out those similar images.In order to improve the accuracy of image retrieval,an improved method is proposed,which uses image visual features as input vector for Support Vector Machine(SVM).Based on the result of image classification and matching of similarity measurement,images from the same class are found out.Experiment results indicate that the method is superior to CBIR.