Abstract: In order to improve the accuracy and speed of image retrieval, a method based on multi-features is presented in this paper. Based?on?the asymmetrical quantization of color space, improved color coherence vectors?are used to extract the?color?feature; improved gray level co-occurrence matrix is utilized as?the texture feature; Krawtchouk moment invariants are introduced to extract the shape feature; it is clustered based on the contribution and establishes image feature index library. The similarity between images is computed based on multi-features fusion, and image fast retrieval is obtainable by using the feature index. Related experiments show that the retrieval precision of the proposed algorithm is improved greatly with faster retrieval speed.

Key words: color coherence vectors, gray level co-occurrence matrix, Krawtchouk moment, contribution

摘要： 为了提高图像检索的准确率和速度，提出了一种多特征组合的图像检索算法。在颜色空间非均匀量化的基础上，利用改进的颜色聚合向量方法提取图像的颜色特征;基于改进的灰度共生矩阵提取纹理特征参数;利用Krawtchouk矩不变量提取图像的形状特征；基于贡献度聚类并建立特征索引库。融合上述特征计算图像间的相似度，使用特征索引对图像进行快速检索。实验结果表明，提出算法的检索精度有较大提高，能快速检索出用户所需的图像。

关键词: 颜色聚合向量, 灰度共生矩阵, Krawtchouk矩, 贡献度