Study Of Face Image Retrieval Techniques Based On Global Semantics And Local Reference

In recent years,with the popularity of electronic devices and the rise of social networking sites and blog,multimedia resources are becoming more and more popular,such as images and video.As the face images have had the largest percentage of all images,how to process and retrieve face images efficiently has become a hot spot in the field of computer vision research.Face image retrieval technology is closely related to image retrieval technology and face recognition technology.Compared to the general image retrieval technology,it tends to focus more on extracting high-dimensional and robust facial semantic feature.Compared to the face recognition technology,it pays more attention to the use of index.Therefore,we need to combine the image retrieval technology and face recognition technology to get the balance of scalability and accuracy of retrieval results.In order to address these issues,this paper proposes the study of face image retrieval techniques based on global semantics and local reference,the main research achievements are summarized as follows:1.We propose a method based on a deep convolution neural network to extract the global semantic feature.With those characteristic,we can build the learning of the hash code and discriminative attributes;2.In order to improve the robustness for factors such as illumination,expression and posture,we make the first attempt to integrate the discriminative binary code learning jointly optimizing the identity-constraints and coding-stability;3.We break up the face into four functional regions,and also get a strategy of selecting reference people,making the choice of local reference people more representative;4.We add the discriminative attributes classification and local reference people classification together to the original binary code,improving the performance of the final face image retrieval.Finally,this article uses the method based on global semantics and local reference for the face image retrieval experiment,and the experimental results show that our proposed method is less time consumption and has the high accuracy,the improvement of face image retrieval performance is significant.