An Image Retrieving System Based on User-Specified Recognition Model

In this paper, we shall discuss an extension of MTDM model for image retrieval based on user's specification. A MTDM model is composed of objects which describes the static and the matching cont,rol methods for recognition targets. Characteristics of the describing targets are declared by abst,ract members of the MTDM model objects. In t,his way, the differences of data description can be casily hidden. In our proposed image retrieval system, user specifies a MTDM model to each keyword for image retrieval that may be not exist in the image database. The retrieval acquisition is structured and represented in the form of MTDM model. When a keyword is not found in the image database, the relat,ed reasoning model(MTDM model) will be fired to match corresponding objects form each image. By the extension of MTDM model, the differences of dat,a are integrated. The image retrieval system can h a t objects in a general way and the recognition model and the retrieving model can be integrated by the MTDM model.