A Noise Reduced FCM-Thresholding Method for Change Detection

Abstract

Change detection is a major application in remote sensing. In this paper, we put forward a novel approach for change detection in synthetic aperture radar images. The approach classifies changed and unchanged regions by fuzzy c-means(FCM) clustering with thresholding. Here we compare the proposed method with MRFFCM which is a modified version of FCM with a novel Markov Random Field (MRF) energy function. Images with speckle noise will results in reducing the contrast of image and is difficult to perform image processing operations like edge detection, segmentation etc.… First, in order to reduce the effect of speckle noise, we introduce a noise reduction technique. A Lee filter is used to remove speckle noise as adaptive filters are more likely to preserve details such as edges or high texture areas. Second, the proposed approach modifies the result of FCM clustering by applying a threshold. In fuzzy c-means clustering the segmented part of the SAR image is not clearly visible. We use global thresholding method and the entire image is segmented using Otsu’s thresholding method. We also compare the results of clustering.