Using the exact same parameters as blurha.sh produces different results

[StainlessStlRat]

Hello! I am using the library and have been for a while, but I've noticed for a while that it produces different (and arguably worse) results than the website implementation at blurha.sh

Please see my reference images. Any idea what could be going wrong?

Here is the generated blur hash and the difference in results:
LTIXH#ai5XRo0xRRnzR-4.M}-$V[

[StainlessStlRat]

NOTE: The website and the encoding in go are using the same number of components on the same axes

[StainlessStlRat]

Reference image for testing:

[StainlessStlRat]

I think there is a subtle bug in multiplyBasisFunction, since in the C source code it's using 2D indexing and the go code is hand rolling the indices and such and other optimizations.

[StainlessStlRat]

Any chance to take a crack at this?

[StainlessStlRat]

Hello, just pinging you again in case you get the itch to take a crack and fix it! I'd love to regenerate all my site's blur hashes with the improved algorithm!