Abstract—This paper presents a new two-stage approach to impulse noise removal for images based on wavelet network (WN) . The first step of this two-stage approach is noise detection , in which the so-called Gray-Level Difference (GD) and Average Background Difference (ABD) is considered as the input of a Wavelet Network (WN) . The second step is removing impulse noise with a median filter . The Wavelet Network presented here is a Fixed Grid Wavelet Network (FGWN) without learning . Experimental results of the study show that our method acts on impulse noise effectively , and at the same time preserves chromaticity and image details very well .

Index Terms—Impulse noise, wavelet network, gray-level difference, average background difference

Amir Reza Sadri is with Isfahan University of Technology, I sfahan (e-mail: ar_sadry@yahoo.com)

Cite: Amir Reza Sadri, Maryam Zekri, Niloofar Ghaisari, and Saeid Sadri, "Impulse Noise Removal from Images: A Wavelet Network Approach," International Journal of Computer and Electrical Engineering vol. 4, no. 6, pp. 882-886, 2012.