Fuzzy Rule-Based Image Processing with Optimization
Kaoru Arakawa · Studies in fuzziness and soft computing · 2000
Fuzzy rule-based image processing technologies for noise reduction and edge extraction are described. Here, two types of noises are considered in noise reduction, namely white Gaussian noise and impulsive noise. Fuzzy rules are applied in order to consider the nonstationarity and uncertainty of signals. Moreover, the fuzzy reasoning part is designed optimally by expressing the system as a nonlinear function of multiple local characteristics of signals, and by setting the nonlinear function so that the mean square error of the output is the minimum for some training image data. Accordingly, the membership function and the rules are automatically designed from this optimization. Computer simulations verify the effective performance of this image processing technology.