Impulse noise removal by k-means clustering identified fuzzy filter: a new approach

Abstract

Removal of impulse noise from corrupted digital images has been a hitch in the field of image processing. Random nature of impulse noise makes the task of noise removal more critical. Different filters have been designed for noise removal purpose and have shown formidable results mostly for low and medium level noise densities. In this paper, a new two-stage technique called k-means clustering identified fuzzy filter (KMCIFF) is proposed for de-noising gray-scale images. KMCIFF consists of a k-Means clustering-based high density impulse noise detection, followed by a fuzzy logic-oriented noise removal mechanism. In the detection process, a 5 × 5 window centering upon each pixel of the image is considered. K-Means clustering is applied on each 5 × 5 window to group the pixels into different clusters to detect whether the central pixel of each window is noisy or not. In the noise removal process, a 7 × 7 window centering upon each noisy pixel of the image, as detected by the clustering is considered. Fuzzy logic is used to find the nonnoisy pixel in each 7 × 7 window having the highest influence on the central noisy pixel of the window. Finally, that pixel is replaced by the approximated pixel intensity value calculated from the highest influencing non-noisy pixel. KMCIFF is evaluated upon seven different standard test images using peak signal to noise ratio (PSNR), structural similarity index measurement (SSIM), Percentage of actual nonnoisy pixels detected as erroneous out of the total number of pixels (PDAE) and average run time (ART). It has been observed that KMCIFF shows significantly more competitive visual and quantitative performances visa -vis most of the extant traditional filters at high noise densities of up to 90 % .

A Neighborhood Impact Driven K-Medoid Clustering and Fuzzy Logic Blended Approach for High Density Impulse Noise Detection and Removal

“…Its pixel intensity value is denoted as PVMNNP. Then PVNP is calculated by utilizing PVMNNP and PVNNPNM using self-published earlier fuzzy oriented approach [48]. The median pixel PVMNNP of the noisy pixel's non-noisy neighbours and the noisy pixel's closest nonnoisy neighbour PVNNPNM are used as reference pixels.…”

“…However, fuzzy logic, on the other hand, is superior to other statistical approaches that rely on comprehensive human understanding of the system when dealing with ambiguous situations characterized by vague and inaccurate facts. To utilize the effectiveness of fuzzy logic, Bandyopadhyay et al [48] employed K-means clustering methodology to identify impulsive noise coupled with a fuzzy logic-based noise reduction strategy. Both of these methods were used in recent times.…”

Clustering with density based initialization and Bhattacharyya based merging

“…When the dimensionality increases, it will be more difficult to verify whether the dataset possesses a clustering structure and to evaluate the results of clustering algorithms. These complexities have led to different clustering algorithms with some restrictions on the shape of clusters such as spiral, hyper-ellipsoidal, etc.. One of the basic and also a popular clustering algorithm is the k-means algorithm [1][2][3]. It is a centroidbased clustering algorithm.…”