基于可微收缩函数与自适应阈值的小波去噪

Wavelet denoising based on differentiable shrinkage function and adaptive threshold

为进一步提升高斯噪声的去噪效果,提出了基于可微收缩函数与自适应阈值的图像去噪方法。根据高斯噪声的小波系数具有幅值小、服从高斯分布的特征,提出一种自适应于信噪强度的阈值,以准确地区分噪声系数与图像系数。根据自然图像的小波系数具有平滑连贯的特征,提出了一种可微的收缩函数,与自适应阈值结合对含噪的小波系数进行量化处理,以有效地去除噪声系数,保持和恢复图像的系数。实验结果证明,相对于现有的最新提出的小波阈值去噪方法,所提出的方法既能更加有效地去除噪声,又能更好地保持和恢复图像的细节和纹理结构。

In order to advance the denoising performance for Gaussian noise, an image denoising method based on differentiable shrinkage function and adaptive threshold is proposed. As per the fact that the wavelet coefficients of Gaussian noise obey Gaussian distribution with small amplitude, a threshold adaptive to the signal and noise intensity is designed, so as to accurately distinguish the noisy coefficients from the image coefficients. And in view of the fact that the wavelet coefficients of natural image have the characteristics of smoothness and continuity, a differentiable shrinkage function is proposed, which is to be integrated with the designed adaptive threshold for quantizing the noisy wavelet coefficient, so as to effectively remove the noise coefficients, preserve and restore the image coefficients. Experiments confirm the fact that compared with the existing wavelet based denoising methods proposed recently, the proposed method removes noise more effectively, and is more capable of preserving and restoring the details and texture structures.