基于贝叶斯估计的杂交小波图像降噪新算法

New image restoration algorithm based on Bayesian hybrid wavelet transform

针对传统图像去噪算法易丢失细节特征、峰值信噪比低等问题,受杂交育种学启发,借鉴遗传算法思路,提出了一种基于贝叶斯估计的杂交小波变换图像去噪算法。该算法以贝叶斯小波去噪后的图像作为父本,维纳滤波处理后图像作为母本进行杂交,对挑选出的个体进行逐代杂交和变异生成子代,将最优子代作为算法的最终解,对其解码还原为去噪后的图像。本算法去噪后的图像峰值信噪比远高于传统算法;去噪后的视觉效果也好于传统方法。实验结果表明该方法不仅能有效消除图像噪声,还能较好地保留图像边缘等细节特征。

In order to improve the drawback of the existing image denoising algorithms which was easily to lose detail features and low peak signal-to-noise ratio（PSNR）,inspired by the cross breeding and learned the idea of generic algorithm,this paper proposed an image denoising method based on hybrid Bayesian wavelet transform.This algorithm took the image denoised by Bayesian wavelet threshold as male parent,and image denoised by Wiener filter as female parent to hybridize and mutate,then offspring could be generated.The offspring which meet the best standards would be decoded to be the optimal solution,and restored to be image as the best result of the algorithm.The denoised image using proposed method had much higher PSNR and better visual quality than conventional methods.The result shows that the proposed method can not only denoise effectively,but also reserve the detail features such as edge