一种改进正态逆高斯分布模型的图像去噪算法*

针对传统去噪算法去除含噪声较大的图像时仍有部分噪声残留的问题,本文基于变换域提出一种改进正态逆高斯分布的图像去噪算法。该算法在非下采样剪切波变换域,利用最优线性插值阈值函数改进正态逆高斯模型作为系数分布模型,对高频子带分解系数进行统计建模,以贝叶斯最大后验概率理论实现图像去噪。实验结果表明对于添加不同标准差的高斯白噪声图像,该算法在有效保留图像细节和纹理信息的同时在峰值信噪比方面优于同类去噪算法。

An improved image denoising algorithm based on normal inverse Gaussian distribution model

Aiming at this problem that the traditional denoising algorithms still exist some residual noise when removing large noise from noisy images. This paper proposes an improved image denoising algorithm based on normal inverse Gaussian model. The algorithm can decompose image into frequency coefficients in the non-subsampled shearlet transform, and use the optimal linear threshold interpolation shrink function to improve normal inverse Gaussian model as the coefficient distribution model. Then the algorithm can build a statistic model using the high frequency subband decomposition coefficients, and finally achieve the noise removal using Bayesian maximum a posterior probability. Experimental results show that corrupted images with additive Gaussian noise over a wide range of noise variance. The proposed method can effectively preserve the image details and texture information, and a state-of-the-art performance in terms of peak signal-to-noise ratio.