基于半监督学习的双线性映射图像检索

An Image Retrieval Method of Double Linear Mapping Based on Semi-supervised Learning

"语义鸿沟"是基于内容图像检索中广泛存在的问题。近年来,人们为减小语义鸿沟开展了许多研究工作,并将半监督学习方法用于其中。目前,多数的检索方法只考虑数据点的结构信息,或关注点集中在低层特征。为了充分利用结构信息缩小低层特征和高层语义之间的语义鸿沟,提出了一种半监督的双映射机器学习图像检索法。该方法在低层特征与标签之间建立双线性映射,最后使用Corel图像库同流行嵌入法进行对比,实验表明所提出的方法在检索过程中可以获得较好的效果,精准率有明显提高。

＂Semantic gap＂ is a widespread problem of content-based image retrieval. In order to reduce the se- mantic gap a lot of research work has been carried out in recent years, and will semi-supervised learning is applied to the field of image retrival. Current research on content-based image the data points＇ structure information is con- sideed merely or low-level features is paid close attention only. To make full use of the structure information and fill the semantic gap between low-level features and high- level semantics, a new image retrieval method is introduced, which is based on semi-supervised machine learning and linear mapping. This method is established double bilinear mapping between low-level features and labels. The method is compared it against the Flexible Manifold Embedding and a significant improvement in terms of accuracy and stability is shown based on a subset of the Corel image gallery.