Intel has released Open Image Denoise, a new open-source CPU-based noise reduction system available on GitHub.
As with the popular OptiX GPU-based Denoise technology from Nvidia, this system uses machine learning techniques to speed up the process of eliminating noise in rendered images.
An AI-controlled hardware-agnostic render CPU-based Denoise program
Open Image Denoise was first presented at Siggraph 2018 as a set of “high-performance, high-quality Denoise filters for images rendered with ray tracing”. It is designed to eliminate the noise created by Monte Carlo rendering methods, such as path tracing which is widely used by modern renderers in production.
As part of the Intel Rendering Framework, Open Image Denoise includes the Embree, a ray tracing kernel library, and the OSPRay, an open-source ray tracing renderer, and works independently on both of them. Unlike OptiX, Open Image Denoise does not depend on the hardware you use: Although it is designed for Intel 64 processors, it supports "compatible architectures", but there is no information on how well it works with AMD processors.
Open Image Denoise in real production
Last year, the original Intel’s demo showed the use of Open Image Denoise with a multi-billion-object island scene from the Disney film Moana as an example.
It works well for production with fewer requirements too, which was proven in the gallery section on Intel’s product website by demonstrating how the new denoiser is used with standard Lumberyard, Crytek and Evermotion test scenes.
On BlenderArtists.org, many artists say that they use Intel's denoiser for rendering interiors and cars, which suggests that the quality of the software’s performance is comparable to the standard denoiser in Blender.
Availability
Open Image Denoise runs on 64-bit Windows, Linux and macOS under the Apache 2.0 license, free of charge. A CPU with support for at least SSE4.2 is required. Both source code and compiled builds are available for free. At this time, the current version is 0.8.0.