Lightnews — Scholar-powered news

Reposted by Pekka Väänänen

Jacco Bikker 🎦

@jbikker.bsky.social

Awesome writeup on the power of pre-splitting for producing high-quality BVHs:
github.com/BoyBaykiller...
One interesting quote: "You may notice PreSplitting improves performance but increases the SAH cost. SAH is not a perfect predictor of performance."

GitHub - BoyBaykiller/IDKEngine: OpenGL Toy-"Engine" with Wavefront Path Tracer and Rasterizer with Voxel Cone Tracing/Global Illumination

OpenGL Toy-"Engine" with Wavefront Path Tracer and Rasterizer with Voxel Cone Tracing/Global Illumination - BoyBaykiller/IDKEngine

github.com

January 29, 2026 at 12:40 PM

Pekka Väänänen

@pekkavaa.bsky.social

I'm starting to get attached to these fast & loose diagram sketches I'm doing for my book. But redrawing them as clean vector art allows easier editing so it's worth the trouble.

The diagram shows a matrix X with labeled points (1–6) representing an original image, and a matrix Y with corresponding labeled points with the name c (sub index 1–2) showing a reconstruction. Below each matrix is a point diagram: the left one for the original image with scattered colored dots divided in two classes, and the right one for the reconstruction with connected dots showing how the two clusters are represented by their mean positions. The formula between them indicates a difference squared, representing reconstruction error.

January 22, 2026 at 8:32 AM

Pekka Väänänen

@pekkavaa.bsky.social

Here's a little project I've been working on: Liero64 - a port of the 1998 MS-DOS game for the Nintendo 64.

Gameplay video: youtu.be/aemLngAjEi8
ROM: codeberg.org/pekkavaa/Lie...

[N64] Liero64 Gameplay

The classic 1998 MS-DOS game Liero is now on the Nintendo 64! This video shows me (left) playing against the CPU opponent (right). Download the ROM at…

youtu.be

January 18, 2026 at 11:17 AM

Reposted by Pekka Väänänen

Makkon

@makkon.bsky.social

screenshots of the startmap I built for qbj3 or Quake Brutalist Jam III
some fancy bits in here will be in the coming update, like the paintings and plaques

a dreary sky, a lonely grey concrete tower in the center. there are chambers on the base that a person can enter with stairs leading up to them

a brutalist concrete lobby, with nice carpet, furniture, and plants hanging from suspended planters. the lighting is warm and welcoming

a cavernous gallery room, with a high ceiling and artificial sun panels pouring warm diffused light into the room. there is a statue of a floating octahedron in the center. plants and vines hang down from the walls. the room branches off into other chambers.

view of a tall, oddly shaped hallway. carpet down the whole length. on the left there is seating, a cabinet, and a large painting of some abstract rock formation.

January 11, 2026 at 9:52 PM

Pekka Väänänen

@pekkavaa.bsky.social

A comparison of the Pillow library's Maximum Coverage color quantizer settings. A palette that works well for dithering may not look great on its own. Here the top result has garish colors but when it's used when dithering (second from top), it's actually quite OK.

Four images of a red building scene, each resulting from different combinations of k-means refinement and dithering settings: top row (k-means, dithering), second row (k-means, dithering), third row (k-means, dithering), bottom row (k-means, dithering).

Figure caption:
Results of Pillow’s MAXCOVERAGE color quantizer (aka the maximin k-means initializer) with varying k-means refinement and dithering settings. The bottom “☒ k-means, ☒ dithering” result may appear the best at first glance but the “☐ k-means, ☒ dithering” (second from the top) image has more vibrant colors when viewed from afar.

January 12, 2026 at 8:32 AM

Reposted by Pekka Väänänen

David Aerne

@meodai.bsky.social

Built a palette extractor that thinks like a curator, not a pixel counter.

– Physics-based color repulsion
– Auto-detects muted vs vibrant bias
– Adaptive spacing to hit target count
– OKLCh color space (perceptually uniform)

Early Access: please break it!

chipper-otter-d6c600.netlify.app

January 8, 2026 at 7:51 PM

Pekka Väänänen

@pekkavaa.bsky.social

Some of you might know that I've been working on a book. Today I'm happy to announce

Palette Programming: A Modern Guide to Color Quantization
paletteprogramming.com

It's a textbook on how to design palettes and represent images using them. Will be out this year. Code in Python & NumPy.

The book cover for Palette Programming: A Modern Guide to Color Quantization by Pekka R. H. Väänänen. The book title and author name are written in black monospace font on a white background at the top part of the cover. The rest of the cover shows a 4x6 grid of images. The images show a photo of a red house getting progressively more colors.

Cover design by https://www.instagram.com/porcelain_hive/

A sample spread from the book's NumPy tutorial. The two pages show prose, Python code snippets, and a color quantization result figure in the top left corner.

January 7, 2026 at 3:13 PM

Reposted by Pekka Väänänen

nicebyte

@nicebyte.bsky.social

a couple 3D aabb tricks:

gpfault.net/posts/aabb-t...

A Couple 3D AABB Tricks

gpfault.net

December 24, 2025 at 7:55 AM

Reposted by Pekka Väänänen

PRBG_Aesthetics

@prbgaesthetics.bsky.social

Ecstatica II (1997)

Developed by Andrew Spencer Studios.

December 21, 2025 at 9:48 AM

Pekka Väänänen

@pekkavaa.bsky.social

I've been reading up on k-means clustering again. I learned that scikit-learn's default "k-means++" init is actually "greedy": it does multiple attempts when randomly picking the next center to add. And the default value for the number of tries is based on an offhand remark in the original paper.

Python code that says:
# Set the number of local seeding trials if none is given
if n_local_trials is None:
# This is what Arthur/Vassilvitskii tried, but did not report
# specific results for other than mentioning in the conclusion
# that it helped.
n_local_trials = 2 + int(np.log(n_clusters))

December 19, 2025 at 8:32 AM

Pekka Väänänen

@pekkavaa.bsky.social

I tried to use premultiplied alpha on the Nintendo 64 but I don't think it's actually possible. The hw blending equation does an unfortunate rounding when mixing colors, adding some black even in fully transparent pixels. Left: Straight alpha, Right: My attempt at premultiplied alpha.

A rotated and alpha blended sprite with a black fringe.

A rotated and alpha blended sprite surrounded by a light grey box. It's an artifact of broken premultiplied alpha blending on the Nintendo 64's RDP.

December 12, 2025 at 8:32 AM

Pekka Väänänen

@pekkavaa.bsky.social

I was looking for how to make deleting files a tiny bit safer in shell scripts and found this subtly terrible StackOverflow answer. Consider the case where $SOMEPATH is empty and what does `cd` do then 🤠

I've found a big danger with rm in bash is that bash usually doesn't stop for errors. That means that:

cd $SOMEPATH
rm -rf *

Is a very dangerous combination if the change directory fails. A safer way would be:

cd $SOMEPATH && rm -rf *

Which will ensure the rf won't run unless you are really in $SOMEPATH. This doesn't protect you from a bad $SOMEPATH but it can be combined with the advice given by others to help make your script safer.

December 9, 2025 at 6:56 PM

Pekka Väänänen

@pekkavaa.bsky.social

I've been reading about color science for my book and I wanted to verify the claim that Oklab gives better linear gradients than the otherwise excellent "CAM16-UCS" color space. And yes, its gradient obviously too desaturated in this test. Oklab's ancestor "IPT" also works fine.

A blue-to-white gradient in the sRGB, CIELAB, CAM16-UCS, Oklab, IPT, and Optimized IPT color spaces. The CIELAB and CAM16-UCS results are clearly off. The gamma-space sRGB interpolation is also slightly too purple.

December 4, 2025 at 5:32 PM

Pekka Väänänen

@pekkavaa.bsky.social

It's lovely when I see dedication to an experiment: keeping a CRT on for a month to avoid affecting its color reproduction. From Fritz Ebner's 1998 dissertation: repository.rit.edu/theses/2858/ The same hue matching experiment data was used by Björn Ottosson when creating the Oklab color space.

"The experiment was performed on a Sun Sparc Ultra 1 with a 20" Sony GDM-20E20 monitor. The resolution was 1152x900 with a refresh rate of 76 Hz and a 0.31mm-phosphor trio pitch. The CRT display was kept on during the extent of the experiment (which ran from May through July, 1997) to minimize drift from power cycling. The layout of the interface is shown in figure 31. The stimulus squares for both test and reference subtended an angle of 4.2 degrees."

November 27, 2025 at 2:00 PM

Pekka Väänänen

@pekkavaa.bsky.social

Here's an unexpectedly brilliant introduction to perceptual color spaces www.youtube.com/watch?v=nJlZ...

A perceptual color space in Minecraft (Oklab)

What is a perceptual uniform color space and how can we use it in Minecraft. Link to Oklab post https://bottosson.github.io/posts/oklab/ World…

www.youtube.com

November 25, 2025 at 6:56 PM

Pekka Väänänen

@pekkavaa.bsky.social

I recently translated my book manuscript (it's about palettes) to Typst and been pretty happy so far. Today instead of writing I procrastinated by moving code snippets' line numbers over the left margin to salvage three extra characters of horizontal space😃

A snippet of my book manuscript that shows paragraphs of text around code snippet.s The line numbers of the code snippets are shifted left over the margin.

November 21, 2025 at 6:56 PM

Pekka Väänänen

@pekkavaa.bsky.social

The Wikipedia page on Color appearance models at en.wikipedia.org/wiki/Color_a... surprisingly good: to the point but still feels comprehensive. Includes a quick comparison of different perceptual color spaces.

Color appearance model - Wikipedia

A color appearance model (CAM) is a mathematical model that seeks to describe the perceptual aspects of human color vision, i.e. viewing conditions under which the appearance of a color does not…

en.wikipedia.org

November 17, 2025 at 8:32 AM

Reposted by Pekka Väänänen

mmalex

@mmalex.bsky.social

via the magic of laion_clap embeddings and umap, my live coding thingy has a sample browser at last!

October 31, 2025 at 6:27 PM

Pekka Väänänen

@pekkavaa.bsky.social

What if Doom's "colormap" was computed in Oklab instead of RGB?

A short article about it on my site: 30fps.net/pages/doom-o...

A Doom screenshot using a slightly more accurate colormap for lighting. Of course differences to the original become apparent only in a side-by-side comparison.

A crop comparison of the earlier image. On the left, the original has bright grey lines in a dark area. In the new version on the right, the lines are gone, making the image more even.

October 30, 2025 at 8:32 AM

Reposted by Pekka Väänänen

Hillsguy

@hillsguy.bsky.social

Silent Hill 4 uses interesting volumetric light / light glare approximation.
Instead of drawing sprites at each light source, full-screen subdivided plane with procedural vertex colors is rendered and blended with game frame.
#gamedev

October 29, 2025 at 12:53 AM

Pekka Väänänen

@pekkavaa.bsky.social

Relatable😑
From: doomwiki.org/wiki/Carmack...
The line in question: github.com/linguica/dmu...

Carmack's typo was a minor bug in dcolors.c, a program written by John Carmack in 1992 during early development of Doom. To generate the invulnerability colormap, Carmack converted each color in the palette to a gray scale value using a standard formula, where the gray scale value equals red*0.299 + green*0.587 + blue*0.114. In the utility, Carmack accidentally wrote 0.144 instead of 0.114, meaning that the invulnerability colormap converts all blues to be slightly too dark.[1] Since the invulnerability colormap inverts the gray scale palette, this means the gray scale colors converted from blue are calculated as brighter than they should be, and so become darker than they should be when the palette is inverted.

October 27, 2025 at 8:32 AM

Pekka Väänänen

@pekkavaa.bsky.social

A new article: Why CIELAB doesn’t improve median cut

It's a common misconception that CIELAB will instantly improve palettes produced by the median cut algorithm. Unfortunately it's not that simple.

📜 30fps.net/pages/median...

It gets pretty technical but I added some context to the intro.

This graph shows how eager each color space is to minimize lightness error. The X-axis represents median cut’s splits as it works its way through clusters until all palette colors have been found. The Y-axis shows the sum of squared lightness errors of the whole image, measured as CIELAB’s L∗.

I believe the low lightness weight is the main reason why median cut doesn’t work better in unscaled CIELAB than in sRGB.

October 24, 2025 at 12:50 PM

Pekka Väänänen

@pekkavaa.bsky.social

Wow I didn't know Python has cute chained comparisons: "e.g., x < y <= z is equivalent to x < y and y <= z, except that y is evaluated only once (but in both cases z is not evaluated at all when x < y is found to be false)."

docs.python.org/3/reference/...

6. Expressions

This chapter explains the meaning of the elements of expressions in Python. Syntax Notes: In this and the following chapters, extended BNF notation will be used to describe syntax, not lexical anal...

docs.python.org

October 23, 2025 at 11:52 AM

Pekka Väänänen

@pekkavaa.bsky.social

A fun look into Source 2's precomputed visibility from a mapper's perspective: www.youtube.com/watch?v=XLT2... It's not a BSP anymore but an octree whose nodes get grouped into clusters. Spoiler: Visleaks can still happen😀