Convolutional Neural Network with Median Layers for Denoising Salt-and-Pepper Contaminations

18 Aug 2019  ·  Luming Liang, Sen Deng, Lionel Gueguen, Mingqiang Wei, Xinming Wu, Jing Qin ·

We propose a deep fully convolutional neural network with a new type of layer, named median layer, to restore images contaminated by the salt-and-pepper (s&p) noise. A median layer simply performs median filtering on all feature channels. By adding this kind of layer into some widely used fully convolutional deep neural networks, we develop an end-to-end network that removes the extremely high-level s&p noise without performing any non-trivial preprocessing tasks, which is different from all the existing literature in s&p noise removal. Experiments show that inserting median layers into a simple fully-convolutional network with the L2 loss significantly boosts the signal-to-noise ratio. Quantitative comparisons testify that our network outperforms the state-of-the-art methods with a limited amount of training data. The source code has been released for public evaluation and use (https://github.com/llmpass/medianDenoise).

PDF Abstract

Datasets


Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Salt-And-Pepper Noise Removal BSD300 Noise Level 30% CNN (Median Layers) PSNR 40.90 # 1
Salt-And-Pepper Noise Removal BSD300 Noise Level 50% CNN (Median Layers) PSNR 37.28 # 1
Salt-And-Pepper Noise Removal BSD300 Noise Level 70% CNN (Median Layers) PSNR 32.4 # 1
Salt-And-Pepper Noise Removal Kodak24 Noise Level 30% CNN (Median Layers) PSNR 36.39 # 1
Salt-And-Pepper Noise Removal Kodak24 Noise Level 50% CNN (Median Layers) PSNR 34.35 # 1
Salt-And-Pepper Noise Removal Kodak24 Noise Level 70% CNN (Median Layers) PSNR 31.56 # 1

Methods


No methods listed for this paper. Add relevant methods here