综合颜色特征的彩色图像检索方法

Image Retrieval Based on Combined Color Features

基于内容的图像检索技术已成为当前的研究热点,文章提出了一种综合利用两种颜色特征进行图像检索的新方法。首先,在变换空间建立色度直方图表示图像的颜色分布特征。为进行图像间的相似性度量,对Swain定义的直方图相似性度量作了改进,为弥补全局直方图不包含颜色空间分布关系的缺点,文章提取了另一种颜色特征,即分块的颜色矩,其距离度量为特征矢量的比值相似度。最后,综合利用上述两个特征对图像进行共同检索。通过对真实图像数据的检索实验表明:综合两种特征检索图像比单一特征检索效果更好。

Contented-based image retrieval becomes a hot research topic.This paper presents a new method for image retrieval using two color features.First,hue histogram is generated which represents the distribution of hue values in an image in transformed space.To address the issue of image matching,it improves the measuring standard defined by Swain.In order to solve the problem of lacking the spatial knowledge of global histogram,it also extracts the other color feature,viz.color moments of partitions.This paper defines the distance between two images based on similarity measures.At last,a new method for image retrieval using combined two above color features is proposed.Retrieval experiments us-ing real color image database are carried out.The results show that the retrieval results obtained from combined-features are superior to that obtained from single-features.