基于图像数据挖掘的有向图模型检索方法

A retrieval method based on directed graph model for image data mining

摘要:

为充分挖掘图像数据信息,提出了一种有向图模型检索方法,结合距离测度初次检索和有向图距离二次检索提高图像检索性能。首先,采用传统的纹理、边缘和颜色特征以及特征之间的欧氏距离测度来进行初次检索,得到一个查询排序列表；在此基础上,结合距离测度与余弦测度设计图像之间的相关测度,在不同的相关测度阈值下构建图像数据集的有向图模型集合；最后,计算有向图距离,据此进行二次检索,降低误检现象。在COREL和ImageCLEF两个数据集上的图像检索实验结果表明,该方法的平均精确度和平均召回率指标高。

Abstract:

To fully dig the image data information, a retrieval method based on directed graph model is proposed, for improving performance of image retrieval by combining first retrieve with distance metric and second retrieve with directed graph distance. First, it executes first retrieve by using traditional features including texture, edge and color and the Euclidean distance among features, and obtains a query sort list; On this basis, it builds directed graph models with different correlation metric thresholds, according to correlation metric that is designed by combining with distance metric and cosine metric; finally, it computes directed graph distance and executes second retrieve, for reducing false retrieval phenomena. The results of image retrieval experiments on COREL and Image CLEF datasets show that, this method has average precision and average recall.