Itsuki Kitano and Yuko Osana
Similarity-based Image Retrieval, Self-Organizing Map, Position Fluctuation
In this research, we propose a similarity-based image retrieval considering position fluctuation by self-organizing map with refractoriness. Most of the conventional image retrieval systems retrieve only images where color and position of objects are similar, and can not realize retrieval considering the position fluctuation. In the proposed system, firstly, the similar images are retrieved based on the color information (natural objects), the color information (artifacts) and spectrum (Search Request 1). If there is no similar image corresponding to the key image or the user can not get enough similar images, the search which permits fluctuation of left-and-right position (Search Request 2) and the search which permits fluctuation of up-and-down position (Search Request 3) are carried out. We carried out a series of computer experiments and confirmed that the proposed system can retrieve similar images considering position fluctuation.
Important Links:
Go Back
