抄録
This paper proposes an image retrieval system which searches a database for images similar to a target imagined by a user. The system uses image features, rather than keywords, and retrieves images by reducing the multidimensional feature space generated by the feature vectors. First, the system presents the user with some sample images having suitable feature vectors and requires the user's interaction to indicate which image is similar to the one he/she has in mind. Then, the information from this interaction is used to appropriately reduce the feature space. This process is continued until the system detects the region where the initial target image is located. Since this method requires no target image and no keywords are given for any of the images in the database, it is very practical. Experimental results show the advantage and efficiency of the proposed system.
