IOSR Journal of Electronics and Communication Engineering (IOSR-JECE) e-ISSN: 2278-2834,p- ISSN: 2278-8735.Volume 14, Issue 1, Ser. II (Jan.-Feb. 2019), PP 26-29 www.iosrjournals.org DOI: 10.9790/2834-1401022629 www.iosrjournals.org 26 | Page A Novel Approach for Image Denoising Using Adaptive Median Filter and Principal Component Analysis Dr. Reecha Sharma 1 , Kamlesh Kaur 2 . Assistant Professor, ECE department Punjabi University, Patiala, India M.Tech Student, Punjabi University, Patiala India Corresponding Author: Dr. Reecha Sharma Abstract: Digital images are corrupted by various types of noises. In this paper a methodology is proposed to eliminate salt and pepper noise from digital grayscale images using adaptive median filter and principal component analysis(PCA).The proposed methodology shows better results for peak signal to noise ratio(PSNR) and mean square error(MSE) in differentiation to existing techniques. The proposed filter works efficiently on high noise density (80% and 90%). The adaptive median filter removes the noise from image and principal component analysis algorithm is further applied to obtain sharp edges and boundaries. Key words: Digital images, denoising, median filter, principal component analysis, PSNR, MSE. --------------------------------------------------------------------------------------------------------------------------------------- Date of Submission: 13-02-2019 Date of acceptance: 28-02-2019 --------------------------------------------------------------------------------------------------------------------------------------- I. Introduction Digital image processing is the field in which digital images are used as input and are processed with the help of digital computer. Digital image is made up of large no. of picture elements known as pixels. Digital image processing field is used to enhance the visual quality of images for better human interpretation. Digital images are effected by various types of noise like Gaussian noise, possoin noise, salt and pepper noise, periodic noise ,exponential noise etc. salt and pepper noise can be produced during acquisition and transmission of digital images[1]. Salt and pepper noise is also known as impulse noise. It consists of random pixels having intensity levels of 0 and 255, spread all over the image. Salt means maximum intensity level i.e. 255 and pepper means minimum intensity level i.e. 0,so it appears as black and white dots on the image. A small quantity of salt and pepper noise can change the appearance of the image to a great extent [2-3].therefore to remove salt and pepper noise from digital images various types of filters have been proposed so far. The linear filters denoised the image but produces blurring effect as well. From the previous research it has been concluded that non-linear filters are best suited for removal of salt and pepper noise because of their better denoising power and edge preservation. Median filter is the most useful non linear median filter [4]. Median filtering is a basic tool for smoothing signals and images [5-8-10]. II. Literature Survey Various types of median filters have been developed so far for removing salt and pepper noise. In the median filter, pixel window is moved on every pixel one by one .every pixel is changed with the median of the window. The window size should be n×n, where n is an odd number [6]. The main limitation of the median filter is that it also processes the non-noisy pixels therefore some important details of image are lost. Another limitation of median filter is that its performance depends upon the window size [7].to overcome the drawback of processing non noisy pixels the various kinds of decision based filter were developed in [7] [8] . These filters only modify the noisy pixels in an image and non noisy pixels are left unchanged. Adaptive median filters in [9][10] have been developed so as to overcome the effect of fixed window size. In adaptive filters variable window size is used. For low noise density small window size is used and for high noise density large window size is used. In [6] selective adaptive median filter is implemented in which only selected noisy pixels are processed and have variable window size.