基于内容的图像检索的一个突出问题是提取整幅图像的全局特征,而用户通常只关注一幅图像中的局部区域.对于如今海量的数据库,提取图像的全局特征使得数据库的信息量变得非常大.这样,从巨大的图像特征库中查找匹配的图像特征时检索准确率将大大降低.针对用户感兴趣的局部区域,提出视觉关键词的概念.一幅图像用若干个视觉关键词替代,这样一幅图像的特征量将大大减小.视觉关键词是用户感兴趣的图像区域,这样使得检索更加具有针对性.实验结果表明检索准确率有明显提高.%The content-based image retrieval has a prominent problem of extracting the global feature of the whole image, however the users usually pay close attention to the local region in an image. For the massive database,the extraction of the image′s global feature makes the information quantity of the database increase greatly,which can reduce the retrieval accuracy while the matched image feature is searched in the enormous image feature database. Aiming at the local region of the users′ re-gion of interest,the concept of visual keyword is put forward,with which an image is replaced by several visual keywords to re-duce the characteristic quantity greatly. The visual keyword is regarded as the image region interested by users to make the re-trieval more targeted. The experimental results show that the retrieval accuracy is improved significantly.
展开▼
