We used two neural-network-based solutions for de-noising, both produce promising results. Additionally, user can specify the noise level to reduce in the second method.
Denoise is implemented using two different neural network models: DRUNet and RIDnet. DRUNet is a state-of-art network that can be used in image denoising, deblurring, and even super-resolution. Along with the input image, a noise level must be supplied to the network in order todetermine the level that denoising will take place. The other network architecture, RIDnet, used four Embedded Atom Models (EAM) to perform denoisewithout needing to specify a noise level to reduce.
See the project report for more details.