Adaptive Fuzzy C-means Algorithm with Spatial Information for Image Segmentation

This paper considers the problem of partitioning noisy images into different regions by fuzzy clustering approach. Based on two fuzzy c-means (FCM) algorithms (FCM S1 and FCM S2), we propose four adaptive algorithms (FCM S11, FCM S12, FCM S21 and FCM S22) which utilize the high correlation of image pixels to increase the algorithms’ robustness to noise. Unlike existing algorithms, our algorithms are free of parameter selection. They can automatically determine the balance between using the standard FCM cost function to segment image and using the local spatial information to eliminate the affect of noise. As a result, these adaptive algorithms are very effective in segmenting noisy images. The experiments demonstrate the effectiveness of our algorithms.