3 ms·
Nice! I guess you'll be the first to publish an encoder for this format. I wonder if Fabrice Bellard deliberately never published his encoder in order to encour
by pierrec 6y ago
Nice! I guess you'll be the first to publish an encoder for this format. I wonder if Fabrice Bellard deliberately never published his encoder in order to encourage curiosity leading to this kind of research. Even then I never expected I would see an analysis with this level of detail, it's almost like a course on compression techniques (but more entertaining).
- a_e_k 6y agoI expect I will be the first, unless someone takes what I've posted here and uses it to create and publish their own before I write up mine (or the original one used for the contest gets published). Still, I've posted a new image file in this format at the end to prove that I've done it. I've corresponded a bit with Fabrice Bellard. He'd indicated that he simply hadn't found the time to clean up his encoder for publication. To be fair, I wouldn't yet consider mine to be publishable. I'll be very curious to compare them if he ever does -- the neat thing about modern lossy codecs is that there tends to be a lot of latitude for the encoders to make choices. I'm glad you enjoyed all the detail. I had worried that perhaps there was too much but I like to be thorough.
- eatingCake 6y ago> I've posted a new image file in this format I couldn't resist the urge to follow along and rewrite the code you've presented in Rust, and what do I find at midnight after finally getting it to work and running your test image through? You got me.
- a_e_k 6y agoExcellent! I was hoping that someone would notice that and get a small chuckle out of it.