基于局部路径特征信息神经网络的图像去噪

图像去噪旨在减少或消除噪声对图像的影响,这一过程往往会有高频细节信息的丢失。为了在去除图像噪声的同时保护图像的边缘信息与纹理细节,本文提出了一种能够连接图像局部路径信息的神经网络,该网络训练完成后可以直接对含噪声图像进行降噪,不需要对图像进行预处理。本文提出的神经网络包括3个部分特征提取层、信息连接模块、信息重建层。信息连接模块是该网络的关键部分,通过残差学习连接局部长路径和局部短路径的特征信息。实验结果表明,经本文处理后的图像在有参考的图像质量评价指标PSNR和SSIM上均有明显提升,PSNR最高可以达到34.87 dB,SSIM可以达到0.87以上;在无参考的图像质量评价指标BRISQUE和NIQE上均有明显下降。本文算法对不同水平、不同种类的算法都有相对较好的效果,且性能优于一般算法,在去噪工作中有一定的实用价值。

Image denoising based on local path feature in formation neural network

Image denoising aims to reduce or eliminate the effects of noise on images.This process often has the loss of high frequency information.In order to protect the edge information and texture details of the image while removing image noise,a convolution neural network which can connect image local path information is proposed.After training,the network can directly denoise the noisy image without preprocessing the image.The neural network proposed in this paper consists of three parts,feature extraction layer,information connection blocks and information reconstruction layer.The information connection blocks are key part of the network,and the mixed feature information of the local long path and the local short path is extracted through residual learning.The experimental results show that the image processed by this algorithm has a significant improvement on the reference image quality evaluation index PSNR and SSIM,the maximum PSNR can reach 34.87 dB,SSIM can reach more than 0.87,and the image quality evaluation index BRISQUE and NIQE without reference have a significant decline.The algorithm in this paper has a relatively good effect on different levels and different kinds of algorithms,and its performance is better than the general algorithm.It has a certain practical value in denoising work.