A Unified Approach for Image Segmentation and Annotation to Storage and Retrieval Based on Content

This paper presents a new approach to formalizing the process of image storage and retrieval based on content considering the information extracted from the image through the segmentation phase and through the annotation phase. We constructed the Attributed Relational Graphs (ARGs) for describing the image contents from the independent domain. The structure of an ARG is made in accordance with a graph grammar which is a mechanism to create the semantics of the content of an image. The instances of the ARGs corresponding to the low level features and to the labels associated with the salient objects extracted from the images are stored in an object oriented database. The native query model is used for the retrieval based on content of the images from the database. Our technique, which combines the image descriptors obtained by an unified approach phases of image segmentation and annotation, has a good time complexity and the experimental results on the image datasets show that the performance of the method is robust.