Defuzzification of discrete objects by optimizing area and perimeter similarity

We present a defuzzification method which produces a crisp digital object starting from a fuzzy digital one, while keeping selected properties of them as similar as possible. Our main focus is on defuzzification based on the invariance of perimeter and area measures while taking into account with the membership values. We perform a similarity optimization procedure using on a region growing approach to obtain a crisp object with the desired properties.