Digital Signal Processing Reference
In-Depth Information
Symmetric Normal Inverse Gaussian and Structural
Similarity Based Image Denoising
Yuanjiang Li
1,2
and Yuehua Li
1
1
School of Electronic Engineering and Optoelectronic Technology,
Nanjing University of Science and Technology, Nanjing, China
hmb4507@njust.edu.cn
2
Institute of Electronic and Information, Jiangsu University of Science and Technology,
Zhenjiang, 212003, China
Lyj@bitai.com
Abstract.
An image denoising method based on symmetric normal inverse
Gaussian (SNIG) model within the framework of non-local means (NLM) is
proposed in this paper. We use Structural Similarity (SSIM) to compute the
value of SSIM between the reference patch and its similar versions, and remove
the dissimilar pixels. Besides, the SNIG model is adopted to adjust the
coefficients of these patches with low SSIM in DT-CWT domain. Experiments
show that the proposed method has the capacity to denoise effectively,
improves the peak signal-to-noise ratio of the image, and keeps better visual
result in edges information reservation as well.
Keywords:
Non local-mean (NLM), image denoising, Symmetric normal
inverse Gaussian, Structural Similarity.
1
Introduction
In the area of image processing, wavelet-based methods have strong impact on
various applications such as denoising, restoration, super-resolution and so on. The
key point of this method is to adjust the coefficient of images in wavelet domain
based on some explicit or fuzzy prior information. The method based on the wavelet
provides good approximations in estimation problems in [1]. In [2], a Gaussian
probability density function (Gaussian PDF) is proposed for updating the image
coefficients. Further, because the discrete wavelet transform (DWT) is not shift-
invariant and lack good directionality, it results in pseudo-Gibbs phenomena in [3].
Besides, the dual-tree complex wavelet transform (DT-CWT) has been proposed for
overcoming the problem, which is employed to decompose the image to seven sub
bands in order to reduce the speckle noise in [4]. However, wavelet-based methods
depend on the accuracy of prior information.
Compared with wavelet methods, non-local methods for image processing are
research hot points due to its simple and model-free advantage in the recent years.
Buades et al [5] presented a non-local means (NLM) denoising method based on the
self-similarity of images in the spatial domain. Its main core is to use the weighted
