基于Boosting学习的图片自动语义标注

图片自动语义标注是基于内容图像检索中很重要且很有挑战性的工作。本文提出了一种基于Boosting学习的图片自动语义标注方法，建立了一个图片语义标注系统BLIR（boosting for linguistic indexing image retrievalsystem）。假设一组具有同一语义的图像能够用一个由一组特征组合而成的视觉模型来表示。2D-MHMM（2维多分辨率隐马尔科夫模型）实际上就是一种颜色和纹理特殊组合的模板。BLIR系统首先生成大量的2D-MHMM模型，然后用Boosting算法来实现关键词与2D-MHMM模型的关联。在一个包含60000张图像的图库上实现并测试了这个系统。结果表明，对这些测试图像，BLIR方法比其他方法具有更高的检索正确率。

Boosting-based Automatic Linguistic Indexing of Pictures

Automatic linguistic indexing of pictures is an important but highly challenging problem for researchers in content-based image retrieval.In this paper,a boosting-based automatic linguistic indexing approach is proposed and a linguistic indexing system called BLIR(Boosting for Linguistic indexing Image Retrieval system) is built.It is assumed that images of same semantic meaning can be represented by a model combined with a group of features.2D-MHMM model is found to be such a template for one special kind of color and texture combination,which corresponds to one cluster in feature space.Thus in BLIR system, a large number of 2D-MHMM models are generated and a boosting algorithm is used to associate keyword with models.The system has been implemented and tested on a photographic image database of about(60 000) images.Results demonstrate the effectiveness of the proposed technique which outperforms other approaches.