TY - EJOU
AU - Jin, Ke
AU - Wang, Shunfeng
TI - Image Denoising Based on the Asymmetric Gaussian Mixture Model
T2 - Journal on Internet of Things
PY - 2020
VL - 2
IS - 1
SN - 2579-0080
AB - In recent years, image restoration has become a huge subject, and
finite hybrid model has been widely used in image denoising because of its easy
modeling and strong explanatory results. The gaussian mixture model is the most
common one. The existing image denoising methods usually assume that each
component of the natural image is subject to the gaussian mixture model (GMM).
However, this approach is not entirely reasonable. It is well known that most
natural images are complex and their distribution is not entirely gaussian. As a
result, there are still many problems that GMM cannot solve. This paper tries to
improve the finite mixture model and introduces the asymmetric gaussian
mixture model into it. Since the asymmetric gaussian mixture model can
simulate the asymmetric distribution on the basis of the gaussian mixture model,
it is more consistent with the natural image data, so the denoising effect of the
natural complex image is better. We carried out image denoising experiments
under different noise scales and types, and found that the asymmetric gaussian
mixture model has better denoising effect and performance.
KW - Gaussian mixture model; asymmetric; EPLL denoising model; image denoising
DO - 10.32604/jiot.2020.09071