【作者】 刘鹏宇； 贾克斌； 王卓峥；

【Author】 LIU Pengyu JIA Kebin Wang Zhuozheng College of Electronic Information & Control Engineering,Beijing University of Technology,Beijing 100022 China

【机构】 北京工业大学电子信息与控制工程学院；

【摘要】 在基于内容的图像检索中,不同的图像特征反映了图像不同侧面的内容。如何有效的组织和利用这些特征,从而提高系统的检索性能是一个值得研究的课题。本文提出了一种综合利用图像颜色、纹理、形状和空间关系四种低层特征的图像检索方法。该方法根据彩色图像的纹理特性,定义了一种图像特征统计量, 通过设计的特征权重分配算子,能够依据图像内容自适应的为上述四种特征分配权重,实现了综合利用图像特征的图像检索。实验结果证明,该方法的检索率优于人工赋予权重的线性加权检索方法。更多还原

【Abstract】 In CBIR,image has various inherent features which reflect its content.How to organize and utilize these features effectively to improve the performance is a valuable research topic.An image retrieval method based on four kinds of image low-level features which are color,texture,shape and spatial relationship features was proposed in this paper.According to image texture characteristic,a kind of image feature statistic was defined,by using feature weight assignment operators designed here,the proposed method can assign weight to the above four features adaptively and realize image retrieval by using combined image features.The experiment results show that the method is more efficiently than simple linear combined low-level visual features of fixed weight.更多还原