| As one of the most basic and important multimedia information carriers in people’s life,image has been widely penetrated in various fields by virtue of rich content and intuitive expression.With the development of science and technology,the number of images is exponential increase.Therefore,the efficient retrieval and indexing of images has become a hotspot in the field of information retrieval both at home and abroad.In this paper,a new method of image retrieval is proposed,which can be used to search and retrieve the relevant images in large quantity and large variety of digital images.In this paper,the image retrieval is based on LDA theme model.Firstly,the LDA theme model is used to train the single-target image database to generate the category theme.When the user searches,the user decides the object contained in the image to be retrieved based on the category theme and retrieves the image containing the similar target.This method is proved to be effective in multi-target image retrieval by designing several experiments.In order to allow users to retrieve images more in line with their search intentions,this paper proposes a feedback method based on topic weights.After the initial retrieval,the user submits the relevant feedback according to the satisfaction degree to the retrieval result.The system analyzes the positive feedback and negative feedback of the feedback of the user.According to the topic probability distribution contained in the feedback image,the subject of the summary topic list is carried on weight adjustment,and then searching again,in order to retrieve more in line with the user retrieval intent of the image. |
