基于粒子群算法的图像相关反馈研究

在浩如烟海的图像库中按用户的需求去搜索图像,传统的做法是基于图像的视觉低层特征进行检索。但是由于视觉低层特征与高层语义之间存在语义鸿沟,因此检索结果往往不尽如人意。研究者发现,将相关反馈技术引入图像检索中则能提高系统的检索性能。提出将粒子群优化算法(Particle Swarm Optimization)1]引入基于内容的图像检索之中,通过用户对检索结果进行评价,系统通过学习来动态修改各特征的权值,从而达到模拟人类的思维,提高图像检索系统的检索性能,并通过实验证明该算法的可行性和有效性。

Study on PSO-based Relevance Feedback Algorithm

The traditional way to search the images in the vast multitude of image database is based on low-level features.There is some misunderstanding called semantic gap between visual low-level features and high-level features,which makes the content-based image retrieval results be unsatisfactory.The researchers found that,relevance feedback is introduced in the human-computer interaction technology which can improve the retrieval performance in the image retrieval.Particle Swarm Optimization algorithm 1] was imported into the relevance feedback on the content-based image retrieval.The system dynamically modified the feature weights through learning,so as to simulate human thinking,and improve the image retrieval system performance.At last,the feasibility and effectiveness of the algorithm in experiments was demonstrated.