Question: 10-bit Colormap Resolution Support

[cecilyen]

I noticed that FSLeyes has an option to use a 1024-step colormap, which I assume corresponds to 10-bit colormap resolution. My setup includes an AArch64 MacBook Pro and macOS 15 with native support for 30-bit color depth (10 bits per RGB channel).

I’d like to clarify:
Does enabling the 1024-color option in FSLeyes actually utilize the 10-bit output path of the GPU/display pipeline? In other words, can FSLeyes display true 10-bit grayscale (or color) images on a 30-bit-capable monitor, assuming the OS, GPU, and display are configured correctly?

This would be especially relevant for viewing smooth gradients in floating-point or high dynamic range MRI data.

Thanks for the great work on FSLeyes!

[pauldmccarthy]

Hi @cecilyen, I'm afraid I don't know the answer to this - is there a way you could determine it empirically?

There is a good chance that the answer is no - FSLeyes uses OpenGL to display image data, and Apple essentially abandoned OpenGL around 2018 - it still works on macOS but has been marked as "deprecated".

At the moment I'm hoping that they continue to support OpenGL in its present "deprecated" state, but if/or when they do remove support, it would probably be several months of work to update FSLeyes to continue to work on macOS.

[cecilyen]

Modern operating systems, including the macOS version I used, along with modern GPUs, support High Dynamic Range (HDR), which can output 10-bit color depth. Many new monitors today are capable of accepting HDR signals and can display either native 10-bit color depth or 8-bit color with Frame Rate Control (FRC).

To test FSLeyes' support for 10-bit color depth, I generated a 1024x1024 NIfTI file where each row contains values from 1 to 1024. I have attached the file here for testing:
gradient_image_1024x1024.nii.gz

However, when I view this attached NIfTI file in FSLeyes, I am unable to perceive a noticeable difference between adjacent grayscale values with my eyes. So, it seems that FSLeyes may not be fully utilizing 10-bit color depth support yet.
 
PS. Versions of the software used
macOS version: 15.3.2
FSLeyes version: 1.13.0
FSL version: 6.0.7.17
OpenGL version: 4.1
OpenGL compatibility: 3.3
OpenGL renderer: Apple M2 Max

[pauldmccarthy]

@cecilyen I am personally unable to differentiate between two adjacent 8 bit greyscale colours (e.g. #aaaaaa versus #ababab), so to be honest I am highly skeptical that 10 bit colour depth would be of any practical use in neuroimaging visualisation.

Practically speaking, if I need to differentiate between two voxels or regions with similar intensity, I will adjust the display range (or brightness+contrast) accordingly.

[cecilyen]

Thank you for your response!

I agree that 10-bit color depth alone may not drastically improve neuroimaging visualization, but it is an essential prerequisite for High Dynamic Range (HDR) rendering. HDR not only enables a broader luminance range—enhancing the brightness of highlights and the depth of shadows—but also improves the visibility of subtle details in both dark and bright regions of images. This could be especially valuable for high-contrast neuroimaging data. While FSLeyes doesn't currently implement HDR, supporting 10-bit framebuffers and LUTs would reduce banding and help lay the foundation for potential HDR workflows in the future, especially if a move to Vulkan-based rendering is planned.

Separately, I believe that adopting modern perceptually uniform and colorblind-friendly colormaps could provide an immediate and practical improvement to data visualization, beyond the default grayscale colormap. This blog post offers a great overview of current best practices in scientific colormap design.

Would the FSLeyes team consider adding support for colormaps such as:
Matlab Parula – Parula Overview
ISMRM Lipari/Navia – Colormap Paper
Matplotlib Viridis / Plasma – Matplotlib Colormaps
Google Turbo – Turbo Colormap Blog
Kitware Fast – Kitware Colormap Announcement
Dave Green Cubehelix – Cubehelix Colormap

I’ve attached a zipped example of the Parula colormap here:
parula.zip

Thanks again for your continued work on FSLeyes—it's a great tool, and these enhancements could make it even better for both clinical and research visualization use cases.

[pauldmccarthy]

Hi @cecilyen, unfortunately the FSLeyes team is just me :) My time is split across quite a few projects, so at the moment I'm only able to do any substantial work on FSLeyes once every few months or so. PRs are more than welcome though, if you have the time or resources.

In theory, adding support for 10 bit depth should only require adjustments to the GL initialisation logic, and the logic for creating off-screen frame buffers (called "render textures" in the code). However, I am not sure how much, if any, coupling there is to the use of [0-255] RGB values in the codebase.

I've tried to make it very straightforward to add additional colour maps to FSLeyes - each colour map just needs to be prepared as a plain-text file with one RGB triplet per line with values in the range [0-1] (more details here). These can be installed on a per-user or per-site basis, and you can also customise the order in which the colour maps are presented in the drop-down menu.

While I'm not particularly opposed to adding more colour maps to FSLeyes, there are already many built-in colour maps, including a collection of perceptually uniform colour maps produced by @CPernet. But feel free to open a PR if you wish - it's just a matter of adding the files into fsleyes/assets/colourmaps/, and adding an entry to fsleyes/assets/colourmaps/order.txt.