基于上下文模型的非抽样小波图像去噪

文章提出了一种新的基于上下文模型的非抽样小波图像去噪方法。与传统正交小波变换不同,非抽样小波在图像分解时不对小波系数进行下采样。分解后的每一个小波系数被模型化为一个广义高斯分布随机变量,应用上下文模型估计每一个小波系数的边缘方差,软阈值函数的构造充分考虑了待阈值化小波系数与其邻域小波系数的相关性,产生空间自适应阈值。通过对加噪图像的实验,可以看出本文方法与其它几种传统去噪方法相比,不仅去噪效果有很大的改进,具有更好的重建视觉效果,而且信噪比也有较为明显的提高。

Nondecimated Wavelet Image Denoising Based on Context Model

A new nondeeimated wavelet transform image denoising method which is based on context model is proposed in this paper. In nondeeimated wavelet transform, the wavelet coefficients are not decimated when the image is decomposed, which is different from traditional orthogonal wavelet transform. After the decomposition, each wavelet coefficient is modeled as a random variable of generalized Gaussian distribution. Context model is applied to estimate edge variance of each wavelet coefficient. The construction of soft-threshold function greatly considers the relativity of being threshold coefficient and its neighbor and then, the spatial adaptive threshold is produced. From experiment results, we can see that compared with other traditional denoising methods, our method can not only greatly improve denoising effect and have better reconstruction visual effect, but also enhances the Signal to Noise Ratio （SNR）.