International Journal of Science and Research (IJSR) ISSN (Online): 2319-7064 Volume 3 Issue 4, April 2014 www.ijsr.net Survey Paper on Different Approaches for Noise Level Estimation and Denoising of an Image Bheem Prasad Ram 1 , Sachi Choudhary 2 1 M. Tech. Scholar 1 , Department of Computer Science and Engineering, School of Engineering & IT, Mats University, Raipur India 2 Assistant Professor, Department of Computer Science and Engineering, School of Engineering & IT, Mats University, Raipur India Abstract: Estimation of noise level in an image is a very important parameter to improve the efficiency of denoising. This article presents different approaches used so far by the researchers for the estimation of blind noise level using statistical and averaging method and denoising of an image. The paper also contains problems in different approaches identified by the survey. Keywords: Patch based Noise level Estimation, Principal Component Analysis, Denoising Sub index terms: Kurtosis, Local PCA, Adaptive PCA, Blind Denoise, Non local- PCA, Singular value Decomposition, Non Local mean with PCA, PCA with Local pixel Group 1. Introduction Image noise is random (not present in the object imaged) variation of brightness or color information in images, and is usually an aspect of electronic noise. Noise is important factor of image processing. Noise randomly present on the image it’s not present objectively. We consider blind noise that has not depended on the parameter. Blind noise has not specific parameter so we analysis sweet -able parameter for blind Noise estimation which can give true noise level. The various noise level estimation categories like filter based noise estimation, selection of block based, and al so on various model of noise model are proposed according to own of interested researcher. Estimation method depends on parameter. There performance depends heavily on accuracy of noise level estimation. Blind noise level estimation is an important part of image processing. The different noise estimation model and de-noise algorithm are estimation to the noise and remove to the noise from image. But these noise estimation and de-noising algorithm still cannot achieve the best performance. These uses of Noise model which is well for single independent to additive white Gaussian noise. There are generally they are classifiable into filter based approach, patch based approach, statistical approach 2. Noise Level Estimation Literature survey described related to this block or patch based noise estimation: Images are decomposed into a number of patches. We can consider an image patch as a rectangular window in the image with size W × W. The patch with the smallest standard deviation among decomposed patches has the least change of intensity. The intensity variation of a homogenous patch is mainly caused by noise. 2.1. Noise level estimation method based on principal component Analysis S.Pyatykh, et al [1]: New noise level estimation method based on principal component Analysis of image blocks. Show that the noise variance can be estimated as the smallest eigenvalue of the image block covariance matrix. It is at least 15 times faster compared with the methods with similar accuracy and it is at least 2 times more Accurate than other methods. Method does not assume the existence of homogeneous areas in the input image; hence it can successfully process images containing only textures. Our experiments show that only stochastic textures those only stochastic textures. It’s near to true noise but not efficient result. Drawback: Patch selection not homogeneous selection .so not stability in result. Over estimate in case weak texture and lower noise level. Underestimate in case rich texture and higher noise level. 2.2. Kurtosis Based Noise Estimation Z. Daniel et al [14]: Natural images are known to have scale invariant statistics. While some earlier studies have reported the kurtosis of marginal band pass filter response distributions to be Constant throughout scales, other studies have reported that The kurtosis values are lower for high frequency filters than For lower frequency ones. They propose a resolution for this discrepancy and suggest that this change in kurtosis values is due to noise present in the image. Then suggest that this effect is consistent with a clean, natural image corrupted by white noise. Those propose a model for this effect, and use it to estimate noise standard deviation in corrupted natural images. Noise estimation is Outperform State of art method. 2.3 Patch based Method Noise Estimation D.-H. Shin,et al [3]: In this paper a patch-based method in which the patches whose standard deviations of intensity close to the minimum standard deviation among decomposed Paper ID: 020131546 618