Effect of Convolutional Based Local Information with Different Distance Measures in FCM Classification

Abstract

Conventional classification procedure assumes that, every pixel consist a single identical class in an image, Fuzzy c-Means (FCM) generally defines membership values for a pixel i.e. any real value between 0 and 1 for each class, in place of enforcing a hard label from among any presumed pure class label set. Fuzzy based FCM classification does not incorporate local spatial information to handle noisy pixels. In this work, Fuzzy local information c-mean (FLICM) and Adaptive Fuzzy local information c-means (ADFLICM) method have been tested to handle noise for remote sensing data classification. These classifiers have been tested with various distance measures. From this work it has been found that all classifiers studied with Canberra distance norm and Fuzziness Factor m=1.1 have given overall best classification accuracy.