Study and application on automatic tagging algorithm of image semantic information

Image annotation is an indispensable step in the process of CBIR (content-based image retrieval). It comprehensively considered both features of the image visual and text messages, which can improve the accuracy of the content-based image retrieval and make image search system more accurate when getting target image. Based on the study about the mainstream technology of the current image mark methods, using CMRM algorithm as machine learning model, this paper realized an image semantic automatic tagging module by combining training methods of image texture characteristics and image retrieval technologies based on color. The merger of this module and early results of CBIR enabled the combination of content-based retrieval and keyword retrieval. It made some improvements to the retrieval performance and narrowed the gap of semantics. Experimental results demonstrated that this project can to a certain extent help users more precisely retrieve to their target images more precise.