二级推理在图像去噪中的应用

摘要

In this paper we research the method of image denoising.For the problem of image blurring of traditional denoising method caused by executing denoising process on all pixels,we proposed a Bayesian decision-based two-level inference model.First,we built a Bayesian decision-based two-level inference model,and obtained the parameters required by Bayesian decision through greyscale histogram of image.After the first inferring,we carried out secondary inference on the obtained classification,and finally acquired the classification of noise and non-noise points.Then we combined the two-level inference model with three kinds of denoising algorithms to test the ability of image denoising.It is verified through experiments that the algorithm proposed in this paper can effectively retain the details of original image to maximum extent under the condition of removing image noise as much as possible,this improves the situation in previous image denoising algorithm that the image blurring becomes rather significant after denoising.%对图像去噪的方法进行了研究，针对传统去噪方法对所有像素点进行去噪处理造成的图像模糊化问题，提出一种基于贝叶斯决策的二级推理模型。首先建立基于贝叶斯决策的二级推理模型，通过图像的灰度直方图，获得贝叶斯决策所需要的参数。首次推理判断之后，对获得的分类进行第二次推理判断，最终获得噪点与非噪点的分类。再将二级推理模型与三种去噪算法进行结合，对图像进行去噪测试。通过实验验证，提出的算法能有效地在尽可能去除图像噪声的情况下，最大程度地保留原图像的细节，改善了以往图像去噪算法中去噪后图像模糊化较为明显的情况。