The article should acknowledge that averaging is better to be performed in a perceptually uniform color space.

Examples: - https://github.com/ClickHouse/ClickHouse/pull/60391/files - https://github.com/ClickHouse/ClickHouse/blob/368c06c5cbb21a...

To me, this is the equivalent of developers assuming everything is 7-bit ASCII because that works in their testing... with US English input only. Sure, they might use UTF-8 under the hood, but they never test with anything other than English so all sorts of things end up broken in subtle and not-so-subtle ways. Right-to-left, word breaking, collation, or something.

"AverageColorRGBA8" for example bakes in the assumption of RGB order, 8-bit, and an assumed linear color space. All three assumptions are regularly violated in real-world apps. For example, HDR formats generally use 10-bit integer or 16-bit float primaries. Some formats use YUV instead of RGB.

> pedestrians like me generally don’t have access to Intel’s’ AMX-enabled hardware. The cost-of-entry for Intel’s Sapphire Rapids chips is pretty steep too. Maybe some day it will be ubiquitous in consumer-hardware and share a similar story as AVX-512.

The cost of entry is not high because the servers are available on AWS. It allowed us to make and test a student project with ClickHouse this year: https://github.com/ClickHouse/ClickHouse/pull/60100

I can't even get a rotozoomer running at 30 FPS.

What is the average color needed for, again?