一种三结构描述子的图像检索方法
Image Retrieval Method Based on Three？？Structure Descriptor

针对基元结构描述子在颜色空间的基础上再提取其他特征,导致偏重对颜色信息的描述而降低了图像检索性能的问题,提出一种应用在HSV颜色空间上的三结构描述子(TSD)的特征提取方法。该方法在HSV颜色空间中分别提取颜色和纹理信息,考虑到了颜色和纹理特征的同等重要性,同时避免了颜色信息的过多干扰;在纹理特征提取中,TSD利用像素间的信息变化来表示局部空间结构信息,解决了传统的局部模式方法忽略对局部结构的空间关系描述的问题,获得了更多的空间结构信息。实验结果表明,该方法在3个图像库Corel？？1000、Corel？？5000和Corel？？10000上的检索准确率分别达到78.08％、38.12％和52.12％,与以往基元结构方法相比,检索准确率得到了提高。
A novel feature extraction method called three？？structure descriptor (TSD) used in HSV color space is proposed to address the problem that the image description methods based on texton pay undue attention to the color description so reduce the performance of image retrieval due to directly extracting other features in color space. The proposed method extracts color and texture information in HSV color space and the same importance of both the color and texture is taken into account while the excessive interference of color information is avoided. TSD uses the information change among pixels to represent the local spatial structure information in texture feature extraction, which solves the problem that the traditional local pattern methods disregard the spatial correlation in local structure and captures more information regarding to spatial structure information. Experimental results show that the retrieval precisions of the proposed method reach 78.08%, 38.12% and 52.12% on the Corel？？1000, Corel？？5000 and Corel？？10000 databases respectively, and its retrieval precisions are higher than those of existing methods based on texton