基于加权距离的新的颜色特征的图像检索算法

A New Color Feature Image Retrieval Algorithm Based on Weighted Distance

针对传统的基于颜色直方图为特征的图像检索算法的不足,提出了将像素的空间平均距离作为颜色特征的新的颜色特征提取算法。为满足不同图像库的特征以及不同用户的需求,在相似度量时引入了加权距离。在实验过程中,选取了以50个国家国旗所组成的图像库进行检索,根据查询图像的颜色特征,利用本文所提出的算法,适当加大红色分量的权重,并在同一坐标系中画出了加权距离与图像检索性能关系的PVR曲线。实验证明,本文提出的算法具有一定的鲁棒性。

In order to overcome the shortcomings of traditional image retrieval algorithm based on color histogram, a new algorithm of color feature extraction is proposed. After using pixel space average distance as the new color feature, In order to meet the characteristics of different image database and the demands of different users, the weighted distance is introduced in measuring similarity. Selected image database mainly consists of 50 national flags in the process of experiment, and retrieved image by using the algorithm proposed in this paper. Appropriate to increase the weight of the red component because of the query image mainly in red, and plot the PVR curves of the relationship between weighted distance and image retrieval performance in the same coordinate system. The experiment proved that The algorithm in this paper has certain robustness.