基于色彩与形状信息特征的图像检索算法研究

Image retrieval based on combining color and shape feature

针对单独用颜色特征并不能很好地表达图像内容的问题,提出了综合利用颜色和形状特征进行图像检索的方法。由于颜色直方图无法表达空间分布信息,因此采用的颜色特征为颜色自相关图,并对色调进行重叠量化。而形状特征采用边缘方向直方图,并对方向进行重叠量化。仿真实验表明,综合利用颜色和形状特征比单独用颜色和形状特征进行图像检索的效果要好,提高了图像检索的查准率。

Color feature does not show content of image well, so an approach of image retrieval by combining color and shape feature is presented. Because color histogram cannot fully expresses the spatial distribution of information, this paper introduces an algorithm about color anto-eorrelogram and overlaps quantitative hue. In addition, shape feature uses edge orientation histograms （EOH） and overlaps quantitative direction. Experiment results show that image retrieval by combining color and shape feature is better than image retrieval that only uses the color or shape feature, and it improves the precision ratio and recall ratio.