Abstract: Noise removing in the medical image is still a challenging task in the research feld especially for the imaging devices like mammogram and ultra-sound. Even though advance scanning technology has been invented, these conventional devices play a vital role in scanning the mammogram breast cancer and fetus images. During the image acquisition itself, these images will get corrupted by the physical interference which appears as a noise in an image and affects its visual quality. In this paper we have applied the various traditional and conventional mean based noise removal techniques for the impulse noise corrupted mammogram breast image and standard benchmark. From this comparative analysis we have found that alpha-trimmed mean flters gives the better result than the other flters in terms of PSNR, SSIM and visual quality. Keywords: Impulse noise, Wiener flter, Noise Reduction, Arithmetic mean flter, Medical image. A Comparative Analysis on Various Noise Reduction Techniques Tested for Medical Images A. Ramya 1 , D. Murugan 2 and T. I. Manish 3 1 Department of Computer Science & Engineering, Manonmaniam Sundaranar University, Tirunelveli, Tamil Nadu, India. Email: ramyaanandan10@gmail.com 2 Department of Computer Science & Engineering, Manonmaniam Sundaranar University, Tirunelveli, Tamil Nadu, India. Email: dhanushkodim@yahoo.com 3 Department of Computer Science & Engineering, Adi Shankara Institute of Engineering & Technology, Ernakulam, Kerala, India. Email: manish.cs@adishankara.ac.in I. IntroductIon Digital images played the vital role in the day to day life of the human. It provides useful information like weather forecasting through satellite cameras, traffc monitoring, medical imaging like X-rays, Medical Resonance Imaging (MRI), Computed Tomography (CT), ultrasound imaging etc., [1]. A traditional crisis in image processing is noise reduction [2]. The noise is characterized by its pattern and by its probabilistic characteristics. There is a wide variety of noise types while we focus on the most important types; they are Poisson noise and salt & pepper noise. A large number of linear and non linear fltering algorithms have been developed to reduce noise from corrupted images to enhance image quality [3]. Noise is present as a result of the electronic circuitry of cameras or in the image transmission period. Salt and Pepper noise is impulse type of noise also called as intensity spikes noise It is generated in the transmission channel. [5]. Due to physical interference also the signal forms the noises such as impulse, Poisson, Gaussian noises (Ramya et. al. 2017) [16]. The main function of flters is to suppress either the high frequencies in the image, that is smoothing the image, or the low frequencies, that is enhancing or detecting edges in the image. A digital flter is a system that performs mathematical operations on a sampled, discrete-time signal to reduce or enhance certain aspects of that signal [4]. Lim et. al. had proposed the super-resolution algorithm that combines Spatially Variant Apodization (SVA) with a modifed geometric mean flter to improve the resolution of Synthetic Aperture Radar (SAR) images and reduce side- lobes simultaneously. This method does not required iterative calculation. This method expands the effective bandwidth and improves the resolution while maintaining the side-lobe performance of SVA. The various parameters are fexibly determined based on SNRs [6]. Ibrahim et. al. had described the impulse flter to detect the noise pixel. The detection stage is done with the intensity values, the pixels are roughly divided into two classes, which are “noise-free pixel” and “noise pixel”. Then, the second stage is to eliminate the impulse noise from the image. In this stage, only the “noise-pixels” are processed. The “noise-free pixels” are copied directly to the output image. The method adaptively changes the size of the median flter based on the number of the “noise-free pixels” in the neighbourhood. For the fltering, only “noise-free pixels” are considered for the fnding of the median value [7]. Nardernejad et. al. had presented the method consists of two stages to flter to de-noise the image, the frst stage consists of a second order nonlinear anisotropic diffusion equation with new neighbouring structure and the second is a relaxed geometric mean flter, which processes the output of nonlinear anisotropic diffusion equation. The proposed algorithm enjoys the beneft of both nonlinear PDE and relaxed geometric mean flter. In addition, this algorithm will not introduce any artefacts, and preserves image details, sharp corners, curved structures and thin lines [8]. Yuan et.al proposed the noise detection-based median flters have been widely applied to impulse noise reduction. However, the number of pixels misclassifed is obviously increased in high noise density. To overcome such drawback, Journal of Network and Information Security 6 (1), June 2018, 18-23 http://www.publishingindia.com/jnis