hdrviz

Data visualization in HDR

About

hdrviz is a minimal Python library (~250 lines) that renders 2D numpy arrays as PQ Rec2020-tagged PNGs for HDR-capable browsers.

If your data has more dynamic range than 8-bit color can show — astrophotography, fluorescence microscopy, fractals, log-magnitude FFTs, density maps — hdrviz lets displays render the better contrast that's actually in the data.

This library is an encoder plus a reference notebook widget. For axes, colorbars, channel mixing, pan/zoom — compose with matplotlib, plotly, or viv.

Install

pip install hdrviz

Quick start

import numpy as np
import hdrviz as hv

data = np.random.RandomState(0).rand(400, 600)
widget = hv.imshow(data, cmap="inferno-hdr", peak_nits=4000)
widget   # display in a notebook

Lower-level API

imshow is a convenience wrapper over a three-step pipeline: apply a colormap to map normalized data to RGB linear-light luminance in cd/m² (nits), run the SMPTE ST 2084 inverse EOTF (the "PQ encoding," via colour-science), and write a PNG with an embedded PQ Rec2020 ICC profile that browsers know how to composite in extended dynamic range.

from hdrviz import hdr_colormap, encode_hdr_png

norm = (arr - arr.min()) / (arr.max() - arr.min())
rgb_nits = hdr_colormap(norm, cmap_name="inferno-hdr", peak_nits=4000)
png_bytes = encode_hdr_png(rgb_nits)   # PQ Rec2020-tagged PNG

PNG bytes are the deliverable. Serve them from a backend, write them to disk, embed them in custom HTML — anywhere that wants "an HDR image from a numpy array."

API surface

What hdrviz isn't

• Plotting framework with axes, ticks, labels, colorbars — compose with matplotlib
• Multi-channel mixing, contrast sliders, pan/zoom, multi-resolution tiling — compose with viv, ideally with HDR PNG tiles encoded by hdrviz
• Animated or video HDR — waits for configureHighDynamicRange() to ship in stable Chromium

Demo notebook

notebook.py is a marimo notebook with introducing the HDR data visualization idea:

• A widget that checks your browser's HDR capabilities
• An interactive Mandelbrot explorer with HDR colormaps and click-to-zoom
• The Horsehead Nebula photographic plate from the astropy tutorials archive
• A fluorescence-microscopy frame from scikit-image.data.cells3d, where the membrane channel's ~150× native dynamic range is the showstopper

marimo edit notebook.py --sandbox

Browser support

Tested and works in Chromium-based browsers. Safari mostly works, but had issues with some images not rendering in HDR.

License

MIT — see LICENSE.