Evaluation of image denoising using fuzziness measures

Concept of fuzzy sets has been incorporated at various stages (e.g., input, output, learning and neuronal level) of network and multi-layered perceptron to handle imprecise, incomplete input data and intractable pattern classes for recognition. Its extension to expert system for rule generation and inferencing has been made along with applications to real life data. This shows how pattern description in terms of properties and membership values can be processed by a neural net for fuzzy and crisp classification. A generalized framework for integration of multilayer perceptron and fuzziness measures has been developed to design an unsupervised system for image denoising. Implementation of fuzzy set theoretic operators using neural networks and the utility of these networks in pattern classification and rule generation have been demonstrated. Various ways of integrating fuzzy set theory and connectionist systems for feature evaluation under both supervised and unsupervised modes have been formulated together with the theoretical analysis.