The Research Of Multiple Features And Indexes In Image Retrieval

Image retrieval is the very important research branch of multimedia technology.Recently,there are two branches of image retrieval,Content-based Image Retrieval(CBIR)and Text-based Image Retrieval(TBIR).On the one hand,we would like use some key word to query images when we employ TBIR for image retrieval.On the other hand,if you query images based on some semantic information of those images,it is named as CBIR.Here,semantic information can be explained as visual information of images,such as color,shape or texture.The Bag-of-Visual-Words(BoVW)model is a mainstream in image retrieval.However,there are two main problems in this model:semantic gap caused by image quantification and insufficient feature discriminative power.Those problems will result in the recognition deflection of machine and human and the depression of the call rate and recognition accuracy.And more and more research institutes and internet companies pay their attention on how to solve those problems and how to optimize the BoVW model.That is,they wish to develop more excellent and effective system of image retrieval for improving the recall rate and recognition accuracy.To solve those two problems,we focus on the research of the multiple features and indexes in image retrieval.This paper proposes to choose multiple features to describe interest points and make evaluations for them so that the better judgment result is obtained.And then,we will generate dictionaries for each features and their Cartesian product.As a result,a coupled index for all images will be built.Finally,we utilize variable-weight for the scores from those three dictionaries to obtain the total score for each image in datasets.By this means,we narrow semantic gap and implement feature fusion.At the same time,we also prove that previous fusion approaches are the special cases of our proposed approach by changing the weights.The experimental results on two public datasets show the benefit of our approach,especially at the aspect of recognition accuracy.