Abstract: A new method combing color shape and spatial feature for efficient image retrieval is proposed in this paper.Color feature is extracted by the histogram，shape feature is extracted by zernike moments and spatial analysis is performed by computing the bar centre and the standard deviation of the distribution of colors.In order to improve the retrieval performance，a relevance feedback based on support vector machines is proposed.It is demonstrated that the image retrieval based on multi-feature is superior than which based on single feature.

Key words: image retrieval, similarity measurement, color histogram, Zernike moments, spatial information

摘要： 提出一种组合颜色、形状和空间信息的图像检索方法，用颜色块的颜色直方图表示图像的颜色特征；使用Zernike矩表示图像的形状特征，通过提取颜色块的质心、分布方差聚散度等特征得到图像的空间特征。为了进一步提高图像检索的精确度，提出一种基于支持向量机的相关反馈方法。实验结果表明，基于组合特征的图像检索方法优于基于单一特征的检索方法。

关键词: 图像检索, 相似度计算, 颜色直方图, Zernike矩, 空间信息