IJSTE - International Journal of Science Technology & Engineering | Volume 1 | Issue 11 | May 2015 ISSN (online): 2349-784X All rights reserved by www.ijste.org 328 A Comparative Study of Recent Image Denoising Techniques R. Jayadevan K. A. Navas Assistant Professor Principal and Professor Department of Electronics and Communication Engineering LBS College of Engineering Kasargode Sreepathy Institute of Management and Technology Anjali Ananthan M. Tech Student Department of Applied Electronics and Communication Thejus Engineering College Abstract This paper presents a review of recent algorithms for noise reduction in images. The techniques discussed here deals with the impulse, multiplicative and Gaussian noise. Image fusion technique is employed as a general model for impulse noise meanwhile the static based median filter is mentioned for the salt and pepper noise reduction. Multiplicative noise reduction techniques involve the application of adaptive windowing along with Lee filtering. Additive White Gaussian Noise reduction made use of a new technique called Sliding double window filtering, which is a frequency domain concept. Fibonacci Fourier Transform is used in this technique. The simulation results and the quantitative analysis show that these techniques possess good edge preserving as well as noise suppression capability. Keywords: AWGN, IMF1, IMF2, Impulse Noise, Multiplicative Noise, Multiplicative Noise, Noise Reduction, SBM, SDWF ________________________________________________________________________________________________________ I. INTRODUCTION Noise Corruption in digital images usually occur during acquisition by camera sensors and transmission in the channel Different types of noise that affect digital images include Additive White Gaussian Noise (AWGN), impulse noise, multiplicative noise etc [1]. Hence, the image denoising is one of the most common and important image processing operation. The impulse noise can be caused by a camera due to the faulty nature of the sensor or during transmission of coded images in a noisy communication channel. The nonlinear-median filter is widely used for impulse noise removal. Most of the median based techniques alter the entire image pixels and hence produces poor quality recovered images. In this paper a novel idea for impulse noise reduction is discussed, which employs the technique called image fusion [2] and a new algorithm using the Statistics Based Median Filter (SBMF) [3] to deal with the salt and pepper noise [1] is mentioned. Multiplicative noise, specifically called as speckle noise [4] usually appears in synthetic aperture radar (SAR) images, and it degrades the quality of images significantly. In this paper, an analysis of recent noise reduction algorithms for different noises is carried out. A brief discussion of a new filtering algorithm based on adaptive windowing [4] and local structure detection [5] is done. The filtering scheme employed in this scheme is Lee filtering [6]. A frequency domain technique for AWGN reduction based on Sliding Double Window Filtering (SDWF) is proposed in [7]. The sliding double window filter contains two window types, the transformed window and the spatial window. For AWGN reduction, several techniques such as averaging and Wiener filtering has already been implemented. The sliding double window algorithm relies on the concept of threshold filtering. The transform used here is the Fibonacci Fourier Transform [7], a modification of the Discrete Fourier Transform. The selection of a suitable noise reduction technique primarily depends on the noise type, its statistics, intensity and the application. So a comparison of different denoising techniques based on these aspects would be very useful. This paper aims to analyze different noise reduction techniques for the three commonly occurring noises in the images, namely impulsive, multiplicative and Gaussian, to determine the suitable methods for each type of noise. The paper provides only the analysis of different types of noises separately. The paper is organized as follows. In section 2 a review of noise reduction techniques for the three types of noise are discussed. In order to compare the performance of the discussed filtering algorithms with the conventional methods, some experimental results and a PSNR versus noise density plot are given in section 3. Finally, this paper is concluded in section 4.