Int. J. Advanced Networking and Applications 873 Volume: 02, Issue: 05, Pages: 873-875 (2011) An Efficient Morphological Salt-and-Pepper Noise Detector *Alok Singh, *Umesh Ghanekar,**Chakresh Kumar, ***Ghanendra Kumar *NIT Kurukshetra, Haryana **ISM Dhanbad,***Sharda University Email: nitkalok.singh07@gmail.com,Chakreshk@gmail.com -----------------------------------------------------------------------ABSTRACT------------------------------------------------------------------ An efficient two stage morphological impulse noise detector is proposed in this paper. The proposed method first identifies the noise pixels by comparing the current pixel with the brightest and the darkest pixels in its working window and then in second stage morphological operations based detector is used to improve the performance of impulse noise detector. Simulation results performed on different images shows better results in terms of the qualitative and quantitative measures of the images. Keywords: - Mathematical morphology, salt-and-pepper noise, image filter. ---------------------------------------------------------------------------------------------------------------------------------------------------------- Date of Submission: 16 September 2010 Date of Acceptance: 26 November 2010 ---------------------------------------------------------------------------------------------------------------------------------------------------------- 1. Introduction In removing salt-and-pepper noise, median filter’s performance is better than the linear filters, though traditional median (MED) filter modify both the corrupted pixels and good pixels in the noisy image and filtering performance is greatly influenced by the filtering window size [1]. In order to avoid distorting good pixels, the switching scheme has been proposed [2]-[9]. These filters perform better than the MED filter by identifying the corrupted pixels using noise detectors before filtering. However these proposed noise detectors will identify some corrupted pixels as the noise-free pixels or misclassify the uncorrupted pixels as the noisy pixels. Consequently, these proposed switching filters will damage some fine details in noisy image or retain some impulses in the filtered image at a high noise density. In this paper, we present a two stage impulse noise detector to realize accurate noise detection for wide range of noise density. 2. Proposed algorithm Here we propose a highly accurate noise detection algorithm for wide range of noise density [up to 90%]. This scheme makes possible a most perfect removal of impulse noise in a more efficient manner by keeping the fine details of the image intact leaving the uncorrupted pixel untouched. Let (, ) f ij be the value of the noisy image at pixel location (i,j). for noise detection purpose first of all we will create a flag image b of image (, ) f ij , and b(i,j) will give the flags value at (i,j) location. Initially all flags values of flag will be set ‘0’ and then for noisy pixel flags will be modified to ‘1’ .Steps of our proposed algorithm are as fallows. Step1. Impose a 7x7 window, which is centered on the current pixel, and find out the maximum value S nax and S min under the window. Step2. The following equation is used for the 1 st stage noise detection max min 1 (, ) (, ) (, ) 0 iffij S or f i j S bij otherwise ì = = ï ï = í ï ï î (1) Step3. Use the following equation for the second stage noise detection and modify the 1 st stage binary flag. (( )• )(, ) (( • ) )(, ) (, ) | (, )| 2 f g g ij f g g ij dij fij ο ο + = - (2) Now, comparing d(i,j) with the predefined threshold T, the noise candidate (i,j) will be re-classified as the noise pixel or noise flag b(i,j) is modified in the following way,. ( ) ( ) 1 , 1 , (, ) 0 bij andd i j T bij otherwise = ≥   =    (3) Step4. Repeat step 2 and step 3 for the each pixel in the image and prepare the noise map b, where 1 denotes the