about

This website provides resources to evaluate and compare the different picture formation algorithm available online. Those kind of algorithm are also referred as:

• display rendering transform
• view transform
• LUT

In few words, picture formation is converting physical quantities to a readable visual representation. The concept can be explored further with:

• https://github.com/sobotka/scise/wiki/Picture-Formation
• https://hg2dc.com/2025/03/13/question-34/
• https://github.com/jedypod/open-display-transform/wiki/doc-introduction

The code source for this website and the image generation can be retrieved at https://github.com/MrLixm/picture-lab-lxm.

Caveats

The workflow and the image presented have the following trade-offs:

• The comparison focus only on the SDR capabilities of all algorithm. Some may be HDR-compatible while other not.
• Some assets were already designed by their author to work with a specific picture-formation; their look are bias against it.
• Some picture-formation were not designed to be applied through OCIO and might be more accurate if used in their proper context.
• Quickly comparing images next to each other will more easily reveal differences that might not be noticeable otherwise. It's best to let your eye adapt to the picture before jumping to a conclusion.

Software

This project is made possible thanks to the following (non-exhaustive) free software:

• Python
• uv
• OpenImageIO
• OpenColorIO
• Numpy
• Jinja
• rawpy
• libraw
• Remix Icon

Assets

Is referred by 'asset' the single image file of a specific scene.

Most of the assets were retrieved from the web where they were kindly made accessible by their author. All rights belong to their respective authors, and they are only shared here for education purposes.

The original image files are not yet shared on the source GitHub repository (for licensing reasons). But it's possible to download them yourself by checking their side-car metadata .json file.

Renderers

A renderer correspond to a color-transform that must be applied on physical pixel data to generate a viewable picture. All renderers are color-managed for the same sRGB target (2.2 power function). Their difference lies in the picture formation algorithm they use.

In terms of workflow, the color transforms are serialized in an OpenColorIO configuration, which will be ingested by OpenImageIO's oiiotool for transforming the pixel data.

Contact

For any feedback/requests, see contact options at https://mrlixm.github.io/pages/contact/