HomeReferenceJWildfire and OIDN Denoiser setup

JWildfire and OIDN Denoiser setup

Support for Intels Open Image Denoiser in JWildfire (notes written by Andreas Maschke)

Table of Contents

    Introducing Intel’s Open Image Denoiser

    The Open Image Denoiser is an library of “high-quality denoising filters for images rendered
    with ray tracing”, developed by Intel. So it it kind of similar to NVidia’s OptiX Denoiser,
    but does not require an GPU from NVidia to operate.

    Cutting down render times in a significant way

    The denoiser allows to cut down render times in a really significant way, especially for
    animations, where many details never would become visible, because they would be removed
    by animation compression. So, by computing them in a lower quality and then denoise the
    result, often leads to similar results like before (computing in higher quality and
    applying animation-compression).

    But, the new feature may also be very intersing for still images containing much noise.
    Some flames need a very large render time in order to become smooth.
    With the denoiser this may become a matter of seconds now.

    To help with this, there is preview-button on the preview area, where you can easily
    make visible the effect of the denoiser and compare it with the raw image.

    Limitations

    Currently, only Windows is supported, but Linux and Mac support could be added later.

    Requirements

    This option currently requires Windows as operating system

    Installation

    There is no additional installation required, everything is included.

    For stability and flexibility reasons, JWildfire utilizes the OIDN command-line denoiser,
    made by Declan Russell[2].


    This makes it possible to execute the denoiser on already rendered flames (by execution
    of the tool via the command-line). You can find the tool inside the Denoiser-folder
    inside the lib-folder of your JWildfire-installation.

    Automatic Self-Test

    JWildfire automatically performs a self-test at startup. Only when the requirements are met,
    the OIDN-Denoiser can be selected and executed.

    You will see this as “OIDN”-option for “AI-based denoiser” at the “Anti-Aliasing/Filter”-tab.
    When you don’t see this option”, but think your system does meet the requirements, please
    see Chapter 4. Troubleshooting.

    Rendering flames using the OIDN Denoiser

    Denoiser Preview

    If your system meets the requirements and you selected “OIDN” for the parameter “AI-based denoiser”
    at the “Anti-Aliasing/Filter”-tab, you will see a small blue button (with the typical Intel-logo-colors)
    in the right upper of the preview area.

    Pressing this button causes the OIDN denoiser applied to the current preview image.
    The left part of the image is kept unchanged, so that you can see the difference.
    There is a light green line to separate the raw (left) and denoised (right) part of the image.

    To achieve better results, you may perform a full preview render before executing the denoiser.
    On the other side, performing it one a very low quality preview will demonstrate you the power
    of the denoising effect.

    The preview also works this way when denoising of the flame is actually disabled.
    See next section to activate denoising for a final render.

    Applying the denoise to final images

    You must select the option “OIDN” for the parameter “AI-based denoiser” at the “Anti-Aliasing/Filter”-tab
    to apply the denoiser when a final image is created. This applies to single-rendering as well
    as to the interactive renderer as well as to batch rendering.


    To generate a preview of the denoising effect, see Chapter 8.1 Denoiser preview.

    Preferences

    Per default the OIDN denoiser is deactivated for new flames. You can overwrite this behaviour
    in the Preferences by changing the property tinaDefaultAIPostDenoiser.

    Rethink your quality settings

    The use of this denoiser may be game-changing, especially for rendering animations.
    You can now try out very low quality settings like 100, 50 or even 5.
    This cuts down rendering times amazingly.

    When creating videos for platforms like Youtube, usually a lossy compression occurs.
    So, it often makes not much sense to compute details in first place which get lost
    when you assemble your video. So, when rendering at low quality and using the denoiser
    to “ramp up quality” you will probably often see not much difference in the final
    result, but it will allow you to be more creative by allowing to try out more ideas
    in the same time.

    Troubleshooting

    When you do not see the OIDN-option for the parameter “AI-based denoiser” at the “Anti-Aliasing/Filter”-tab,
    but think your system does meet the requirements, please ferform the following steps:

    1. open a windows explorer and locate the folder of your JWildfire installation,
      e.g. D:j-wildfire-6.00
    2. copy the path into your clipboard
    3. press + to execute a command
    4. type “cmd” into the field and press
    5. enter the command “cd /d “, so that there now appears something
      like “cd /D D:j-wildfire-6.00”
    6. Press enter
    7. Enter the command: “java -jar libj-wildfire.jar >diagnostics.txt”
    8. Press enter, JWildfire should open
    9. Close JWildfire, go back to the console
    10. Enter the command “notepad diagnostics.txt”
    11. Press and look for an error message
    12. Try to resolve it, e. g. by installing the latest driver for your graphics card
    13. When you can not resolve it, report the error, but please include the file diagnostics.txt
    14. Good luck!

    References

    [1] Intel Open Image Denoise: https://www.openimagedenoise.org/
    [2] Commandline Denoiser by Declan Russell: https://declanrussell.com/portfolio/intel-open-image-denoiser-2/

    You may be interested in ...

    Leave a Comment