A COMBINED NEURO-FUZZY APPROACH FOR CLASSIFYING IMAGE PIXELS IN MEDICAL APPLICATIONS

Rami J. Oweis - Muna J. Sunna'

   This paper is concerned with classifying image pixels into three sets of pixels: contour, regular, and texture. When properly processed, classified images can represent foundations for diagnostic purposes. A neuro-fuzzy approach was used to take advantage of neural network's ability to learn, and membership degrees and functions of fuzzy logic, respectively. The method is based on the spatial properties of the image features and makes use of multi-scaled representations of the image. A training set was used to create and train the classifier system. The classes were represented as fuzzy sets with degrees of memberships. Each pixel was assigned a degree of membership for each of the three fuzzy subsets. Classified pixels were finally shown as three separate images each representing a set. The method showed high quality classification for images of simple components. This approach would be highly attractive in the biomedical field due to the vast availability of images.

Keywords: image processing, biomedical images, pixel classification, pixel classes, neuro-fuzzy approach

[full-paper]