Hi, QOI is a really exciting and interesting image format. Thanks for sharing your inspired code.
This update adds a --forcebpp # option to qoibench to allow the number of bytes per pixel to be forced to 3,4,etc. Also provides better logging on to QOI mismatches.
I do machine vision, so speed is the most critical feature in lossless image compression. In that domain the majority of images are still greyscale, so I was inspired to come up with a greyscale version to that would complement your colour one.
Would you consider adding this greyscale implementation into QOI? The code and benchmarks results against the benchmark suite are in the repo above (for 1,3 and 4 byte per pixel encodings).
Hi,
would also see great benefits for the GRAYSCALE variant and a huge amount of applications. I'm curious about the pros and cons. THX for your all superior idea's and implementations.
Hi, QOI is a really exciting and interesting image format. Thanks for sharing your inspired code.
This update adds a --forcebpp # option to qoibench to allow the number of bytes per pixel to be forced to 3,4,etc. Also provides better logging on to QOI mismatches.
This change was made in development of on a version of a complementary QOI-inspired 1 byte per pixel greyscale encoding. It is here: https://github.com/jstavats/qoi/tree/adding_grey
I do machine vision, so speed is the most critical feature in lossless image compression. In that domain the majority of images are still greyscale, so I was inspired to come up with a greyscale version to that would complement your colour one.
Would you consider adding this greyscale implementation into QOI? The code and benchmarks results against the benchmark suite are in the repo above (for 1,3 and 4 byte per pixel encodings).
Keep up the awesome work!
Thanks, Jay