Activity Structyures Based Framework for Image Segmentation
Sabah M.A. Mohammed
Jinan A.W. Fiadhi
Despite the numerous work in image segmentation many people are still looking for a segmentation methodology that can succeed with general images. Activity Structures proved to be successful in designing various knowledge-based systems. This paper focuses of this methodology for image segmentation. It starts by introducing the modeling formalism and showing that it can be used to model linear systems such as image segmentation whose coefficients are rational. Colors are represented as fuzzy sets and any operation on such colors cab be then be achieved via the fuzzy disjunction operators. Image regions are represented as fuzzy relations and operations on their similarities and differences can be achieved via the BK-Relational products. The image segmentation system is considered as a reasoning system that can analyse its performance and make necessary changes to their behaviour in order to satisfy certain predefined control goals. Four types of hybrid controllers have been considered for the physical representation of the reasoning system. Every hybrid controller contains a fuzzy classifier and a self-organizing neural network.
Cited References Fulltext