DefinePK

Abstract

In an image retrieval system, the major challenges are image understanding and image annotation. This paper presents an Interactive Segmentation and Fuzzy Rules (ISFR) framework that includes two major components: Interactive Image Segmentation (IIS) for image understanding and Fuzzy Annotation Theory (FAT) for image annotation. IIS extracts multiple objects and background regions from the images and identifies the user context objects in the images through multiple markers. FAT is used to develop the image annotation systems. The colour, shape, and texture of objects are used to represent the visual concepts of the images. This paper extends and enriches the fuzzy based knowledge representation to map visual concepts to high level concepts. Thus the formal specifications of the visual concepts to the corresponding high level concepts are constructed. The proposed ISFR framework produces automatic annotation of segmented objects and retrieves images that best match user’s expectations. Experiments on benchmark dataset validated that the proposed ISFR framework yields better segmentation than the existing algorithms.