Denoising image of single precision & many pixels using Deep learning

I'd like to denoise a single precision image using Deep learning, which is a provided as "DnCNN" and "denoiseImage".

Noise follows typical Gaussian, and my original image size is about 2000 x 2000, and tif version.

To train the network with my data, I use "imageDatastore and denoisingImageDatastore" as shown in https://www.mathworks.com/help/images/train-and-apply-denoising-neural-networks.html.

• I think I need to use "augmentedImageDatastore" due to my image size over 256 x 256, but it seems impossible using it with "denoisingImageDatastore". Is it right? or there is other way to use them together?
• Another I wonder is... the single precision image should be scaled by 10e-6 to put Gaussian nosie using "imnoise", and I think there is no ways to do that when using "denoisingImageDatastore". Is also right?

Thanks for your help.

Answer 1

[yanqi liu]

yes，sir，may be use denoisingImageDatastore to make data

during process，the data may be normalized before train