Film grain is the visual texture which is due to the use of silver halide crystals in the emulsions of analog films. It is desirable to synthesise this in digital images for artistic and compression purposes. Here are the works I have carried out on this subject.
The goal of this work, which was the PhD work of Gwilherm Lesné , carried out at Télécom Paris (Yann Gousseau, Saïd Ladjal) was to carry out analog film grain synthesis in digital images using a deep neural network. Here is a result of such a synthesis :
Here is a link to the preprint of this work :
The goal of this work, carried out during my postdoc at Université Paris Descartes (now Paris Cité), under the supervision of Julie Delon and Bruno Galerne, was to propose a physically realistic stochastic model for analog film grain synthesis in digital images. Here is the result of such a synthesis :
The paper describing this work can be downloaded here :
Alasdair Newson, Julie Delon, Bruno Galerne, A Stochastic Film Grain Model for Resolution-Independent Rendering, Comput. Graph. Forum36(8): 684-699 (2017)
Alasdair Newson, Noura Faraj, Bruno Galerne, Julie Delon, Realistic Film Grain Rendering. IPOL Journal 7: 165-183 (2017)
We have two versions of the source code, with GPU and without GPU.
With GPU
FIlm grain rendering code (with GPU)
Without GPU (this link contains an online demo and the source code) :
Film grain rendering code (without GPU, IPOL implémentation)
If you wish to use the previous version of the code, please cite our paper :
Realistic Film Grain Rendering
A. Newson, N. Faraj, B. Galerne, J. Delon
IPOL Journal 7: 165-183 (2017)