增量学习语义属性的图像内容检索系统增强

大规模图像内容检索是实现图像语义信息获取的重要手段, 其首要需解决图像低层特征与用户高层语义间的语义鸿沟问题。针对该问题, 引入图像语义属性, 并结合增量分类学习方法(online core vector machine, OCVM), 提出了一种增量构建大规模图像内容检索系统的新方法。该方法借助检索反馈学习机制可以提升图像语义属性的辨别准确性, 能在扩张图像库规模的同时, 提升图像内容检索的可靠性。实验结果表明了上述方法的有效性, 其检索性能可逐步地达到离线构建方法的最佳性能, 但具有更好的可扩展性和自提升能力。

Improvement of content-based image retrieval system byincrementally learning semantic attributes of images

The large-scale content-based image retrieval system must be an important means of getting semantic information, firstly it needed to solve the problem of the semantic gap which between the high-level semantics the users needed and the low-level features of the image. Facing to this question, it introduced the image semantic attributes, and combined with incremental learning method OCVM, this paper proposed a new method which based on building a new large-scale content-base image retrieval system incrementally. It used the retrieval feedback learning mechanisms to enhance the accuracy on identifying the image semantic attributes. With the expansion of the scale of the image library, it could improve the reliability of the content-based image retrieval at the same time. The experimental results show the effectiveness of the above method, and the retrieval performance of the method can gradually achieve the method which constructed offline, but has better scalability and improving capacity itself.