A NEW APPROACH ON NOISE ESTIMATION OF IMAGES

Mredhula L, Dorairangaswamy M A

Abstract: This paper proposes a new idea towards the researches on noise estimation on images. Image Denoising is always a challenging field as visual quality factor plays an important role. The noise estimation process is done in the transformed domain. In this work, Wavelet Transform is used because of its sparse nature. Then a Bayesian Approach is adopted by imposing an a priori Gaussian Distribution on the transformed pixels. The image quality is checked before entering to noise estimation process by using Maximum-Likelihood Decision criterion. Then a new bound based estimation process is designed by taking the idea from Cramer-Rao Lower Bound for signals in Additive White Gaussian Noise. A visually better result is observed in the experimental output which is obtained after reconstructing the original image

Keywords: Image Denoising; Wavelet Transform; Bayesian Approach; Gaussian Distribution; Maximum-Likelihood Decision Criterion, Cramer-Rao Lower bound

DOI: https://doi.org/10.15623/ijret.2014.0327004