中文摘要：

针对图像检索问题,提出一种基于像素差分基元矩阵的图像检索方法。该方法结合图像的颜色特征与纹理特征,在量化后的HSV颜色空间中提出10种基元;通过定义的基元扫描图像,生成像素差分矩阵以及基元过渡矩阵。最后利用统计算法将上述两个矩阵合并为一个像素差分基元矩阵,实现了颜色、纹理以及空间信息等多特征的图像检索。在Corel标准图像数据库中执行图像检索方法间的对比实验,在Corel-4000图像数据库中执行旋转图像检索实验,实验结果表明,该方法不仅具有良好的检索表现,而且可以实现旋转图像的检索。

英文摘要：

In the view of the issue of image retrieval,this paper proposed a new image retrieval algorithm based on a difference between pixels motif matrix（ DBPMM）. Firstly,the algorithm approved ten kinds of motifs combined with the color feature and texture feature of image in the modified HSV color space,and then gained a difference between pixels matrix（ DBPM） and a motif excessive matrix（ MEM） by scanning the image through the special motifs. Finally,it used the statistical algorithm to merge above two matrices into a DBPMM,which could realize the multi features,i. e. color,texture and spatial information,image retrieval. Experimental comparison between image retrieval methods on the benchmark Corel image dataset and the experiments of rotated image retrieval on the Corel-4000 image dataset show that this approach not only has better retrieval performance but also implements the rotated image retrieval.