AN UNSYMMETRICAL TRIMMED MEDIAN USING DISTANCE MEASURE AND MIDPOINT FILTER FOR THE REMOVAL OF SALT AND PEPPER NOISE

Image denoising was a preprocessing step in image processing used to remove the noise while retaining as much as possible the important features. The fluctuations in the pixel value caused by the unwanted disturbance were known as noise. Hence, noise reduction techniques have to be used to improve the quality of the image. This article reports on the results of a research algorithm that uses median, midpoint and mean for the reduction of salt and pepper noise from the grayscale images. The median value was calculated for the uncorrupted pixels which are having the minimum distance. The proposed algorithm was tested using Euclidean distance, D4 distance and D8 distance as the distance measures. In this approach, the 3x3 window was selected as initial window size and the window size was made adaptive based on the noise density. The proposed filter was experimented using standard Lena image, 200 iris images, sample Plant images, sample MRI images and sample CT images in JPG/JPEG (Joint Photographic Experts Group) format. These grayscale images used as a test image were induced with salt and pepper noise density ranging from 10% to 90%. The simulated results proved that the proposed approach performs better than the MDBUTMF, AMMF, MNF, IUTMMF, and HYBRID filtering algorithms reported in the literature in terms of PSNR, IEF and MSE.