High Density Impulse Noise Detection using Fuzzy C-means Algorithm

Abstract

A new technique for detecting the high density impulse noise from corrupted images using Fuzzy C-means algorithm is proposed. The algorithm is iterative in nature and preserves more image details in high noise environment. Fuzzy C-means is initially used to cluster the image data. The application of Fuzzy C-means algorithm in the detection phase provides an optimum classification of noisy data and uncorrupted image data so that the pictorial information remains well preserved. Experimental results show that the proposed algorithm significantly outperforms existing well-known techniques. Results show that with the increase in percentage of noise density, the performance of the algorithm is not degraded. Furthermore, the varying window size in the two detection stages provides more efficient results in terms of low false alarm rate and miss detection rate. The simple structure of the algorithm to detect impulse noise makes it useful for various applications like satellite imaging, remote sensing, medical imaging diagnosis and military survillance. After the efficient detection of noise, the existing filtering techniques can be used for the removal of noise.