Paper notes-Image Denoising via CNNs: An Adversarial Approach

Paper notes-Image Denoising via CNNs: An Adversarial Approach

1.Main task

    Is it possible to recover an image  from its noisy version using convolutional neural networks?The paper propose a new CNN architecture for blind image denoising which synergically combines three architecture  compoents,a multi-scale feature extraction layer which helps in reducing the effect of noise feature maps.

    Image denoising is a fundamental image processing problem whose objective is to remove the noise while preserving the original image structure.Traditional denoising algorithms are given some information about the noise,but the problem of blind image denoising involves computing the denoised image from noisy one with out any knowledge of the noise.

    This paper addresses how CNNs can be used for blind image denoising.They are noisy images  and require the network to gather enought features from this images so that a noise-free version can be computed from them.

    1,They propose a multi-scale adaptive  CNN architecture which gives a competitive performance to the state-of-the-art image denoising approaches.

    2,A train regime which exploits clean images as well as noisy images to get good feature maps for recongstruction

2,background

3,Main solution