Deep convolutional neural network for mixed random impulse and Gaussian noise reduction in digital images
Deep convolutional neural network for mixed random impulse and Gaussian noise reduction in digital images
复制标题
DOI：
10.1049/iet-ipr.2019.0931
复制
发表时间：
2020-12-15
影响因子：
2.3
通讯作者：
Adjouadi, Malek
中科院分区：
文献类型：
--
作者：
Mafi, Mehdi;Izquierdo, Walter;Adjouadi, Malek
本研究利用深度卷积神经网络 (CNN) 实现正则化和批量归一化，以去除数字图像中各种级别的混合噪声、随机噪声、脉冲噪声和高斯噪声。这种深度 CNN 实现了细节损失最小化，并且在处理已知和未知噪声混合物时产生结构指标的最佳估计。此外，通过使用不同结构指标对去噪滤波器进行了全面比较，以突出所提出方法的优点。最佳去噪结果是通过使用具有 40 x 40 块的 20 层网络获得的，该网络在 Berkeley 分割数据集 (BSD) 的 400 180 x 180 图像上进行训练，并在 BSD100 数据集和研究界普遍感兴趣的另外 12 个图像上进行测试。比较结果证明了所提出的滤波器的优点，并且结果的综合评估突出了这种基于 CNN 的方法的新颖性和性能。
This study utilises a deep convolutional neural network (CNN) implementing regularisation and batch normalisation for the removal of mixed, random, impulse, and Gaussian noise of various levels from digital images. This deep CNN achieves minimal loss of detail and yet yields an optimal estimation of structural metrics when dealing with both known and unknown noise mixtures. Moreover, a comprehensive comparison of denoising filters through the use of different structural metrics is provided to highlight the merits of the proposed approach. Optimal denoising results were obtained by using a 20-layer network with 40 x 40 patches trained on 400 180 x 180 images from the Berkeley segmentation data set (BSD) and tested on the BSD100 data set and an additional 12 images of general interest to the research community. The comparative results provide credence to the merits of the proposed filter and the comprehensive assessment of results highlights the novelty and performance of this CNN-based approach.