image-rs / color_quant Goto Github PK

The complex but stable (API unlikely to change) algorithms in math::nq seems to be good subjects of benchmarking. Benchmarks will help those who want to improve their performance.

There is a CUDA implementation of NeuQuant, CuNeuQuant, in C++
(Source code was Exported to this GitHub repository)

Though it is licensed under LGPL-3,

Dynamically linking LGPL-3 library requires no GPL license

arrayfire This crate(under BSD) supports both OpenCL and CUDA for parallel computing

The paper about CuNeuQuant

Example link

colors should be $>=64$

Hello,

I'm using this library and I've recently encountered an issue with a 500x500 image. It seems that it's caused by the size of my image, because scaling down to 300x300 fix the issue.

You can see more explanation here https://stackoverflow.com/a/33723556

Have you ever encountered this issue?

Thank you

There is only 1 algorithm available and that too written specifically for RGBA format, It would be great to have the other algorithms like:

• Uniform quantisation
• K-means quantisation
• Median cut algorithm
• Octree quantisation
• K-d tree quantisation(Minimum variance partitioning)

I can help with last three fast ones

Currently this crate has no tests and only 1 example. I think it would be better to have more of them.

This crate is a dependency of the gif crate and seems to have been forgotten in the move of all image-related crates to the image-rs organization. See image-rs/image#881. It would be a good idea to also move this repository as well as the crates.io ownership to the image-rs organization.

cc @bvssvni @HeroicKatora @fintelia

@nwin Same as the other ones.