E-Governance – Current Scenario 1 A Comparative Study on Hermite Interpolation and other Median Filtering Techniques for the Impulse Noise Removal *Saumya Satpathy, **Figlu Mohanty, ***Prasant Kumar Pattnaik * & ** & ***School of Computer Engineering, KIIT UNIVERSITY, Bhubaneswar, India Abstract The digital images are more likely used in many applications such as medical science, millitary, aero-space and industry. Yet, the images are often corrupted by impulse noise during its acquisition and transmission process. Median filters are the most simple and practical solution to shed impulse noise from digital images. However, some crucial edge information gets deprived in median filtering. Our paper uses Hermite Interpolation in order to preserve edge. The outcome shows a better PSNR value in comparison with existing Median Filtering techniques. Keywords: Hermite interpolation, Digital images, Edge preservation, Impulse noise, Median filter 1. Introduction In real scenario, digital images captured by digital cameras and sensors are mostly degraded by sources (low sensor quality, defects in lens, storage etc). Apart from sensors, noise can be introduced in the digital images during acquisition and transmission which is called as Impulsive Noise. This noise can be categorized as Salt and pepper noise and Random valued noise. The Median Filter is the simplest and widely used mechanism to remove impulse noise from digital images. In [1], the author explained a method consisting of noise detection proceeded by the removal of detected noise by Median Filter using particular pixels that are noise free. The detection of noise depends on thresholding of pixel values. However, the drawback of this paper is the threshold used in the mechanism is analysed every time which is monotonous and time- consuming. Deka, B. [2], proposed a switching based Median Filter for the removal of impulse noise in gray scale images. The idea of the filter is to substitute the irregular pixels first, instead of estimation. The mechanism includes two stages: 1) it comprises of detection of irregular pixels by using signal dependant rank-order mean filter. 2) The irregular pixels are replaced with first order 2D non-casual linear prediction technique and then substituted by the median value. It refines images up to 40% noise density. Nooshyar, M.[5], the whole image is examined and operated pixel-by-pixel in order to get the noisy pixel. A two-dimensional window is created making the noisy pixel as the pivot pixel. According to the number of noise free pixel in the neighbourhood, the window size differs. The revised value of noisy pixel is the outcome after applying Median Filter and Weighted Mean Filter on noise free pixels within the corresponding window. Sable [8], suggested a double bi-lateral filter which is an advanced version of classical bi- lateral filter. This paper proposes a modified double bi-lateral filter where a new decision based algorithm based Median Filter in second bi-lateral replaces 3*3 Median Filter, restoring up to 80% noise density. Sharma, A.[9], the author described a novel method consisting of two thresholds (minimum and maximum) which identify the noise. The enhanced median value substitutes the irregular pixels recursively. Instead of using static threshold values the adaptive dual threshold method uses threshold value that changes dynamically according to the pixels of individual window. 2. Critics on Hermite Interpolation over some Median Filtering techniques We have used Hermite Interpolation for the removal of impulse noise. Hermite curves are simple and user-friendly to determine as well as performs smooth interpolation between key-points. We have used following stages name Stage-I and Stage-II while calculating the