摘要
针对如何在图像检索系统中客观地表达用户的感知 ,提出了一种基于粗糙集理论的相关反馈算法。通过相关反馈过程将用户感知与图像特征相结合 ,利用粗糙集理论归纳用户感兴趣的图像语义特征 ,并根据用户感兴趣的程度调整对应图像特征权重。作者建立了一个实验系统 ISS,采用颜色直方图与语义特征作为图像特征 ,并实现 MARS的反馈算法作为性能比较算法。实验结果表明 ,该算法较 MARS系统在检索性能上有较大的提高。
A novel relevance feedback algorithm for image retrieval is proposed to express human perception objectively. The algorithm combines the human perception with the image feature through the process of relevance feedback and concludes interesting image feature for users. The weights of these interesting image features are also adjusted according to the user perception. An image retrieval system-ISS is constructed. The image features are composed of the color and semantic features. The relevance feedback algorithm for multimedia analysis and retrieval system (MARS) is also implemented in the system to compare with the algorithm. Experimental results show that the retrieval performance of the algorithm is better than that of MARS.
出处
《数据采集与处理》
CSCD
2004年第3期278-281,共4页
Journal of Data Acquisition and Processing
基金
国家自然科学基金 ( 60 1135 0 2 0 F F0 30 40 5 )资助项目
科技部创新基金 ( 0 1C2 62 2 42 10 70 8)资助项目
