does anyone know of such a thing
There was a discussion a while back on Apionet where the conclusion was reached that lossless video compression requires an end to class inequity to work.
the conclusion was reached that lossless video compression requires an end to class inequity to work
no, the conclusion was that lossless video compression is ineffective and pointless for most use-cases. but that doesn't even relate to the question.
I'm guessing the intended meaning of the question is, "is there a reasonably effective video codec that is simple to implement", and I would like to know this also.
but it is true that there could be a use case for lossless video post-singularity
an end to class inequity is not a sci-fi situation where we have infinite storage space
are you not satisfied with gollark's answer? you might want to make your criteria more clear if you want a more focused discussion.
ubq323: Okay, i won't derail it, but i'll reroute the train once the stop has been reached.
Lossless QOIlike video encoding is just QOI on each frame, I think. Maybe you could be slightly smarter but interframe coding is rather tricky.
gollark: it would not be that tricky or complicated to subtract the previous frame from the current one and replace all runs of zeros in the difference with runs of alpha=0 pixels in the frame to be encoded.
Better compression could be achieved with a slightly more complicated scheme, such as computing the total distance between consecutive frames and using a different "delta QOI encoding" for frames where the distance is below a certain threshold. This would have its own separate color cache for INDEX, and DIFF and LUMA would be modified to encode the difference computed with the previous frame rather than the previous pixel in the same frame.
But back to the original question:
One of the main properties of QOI that makes it so simple is that it is streamable. Once a command has been computed, it can just be output because pixels at the end of the image cannot affect how pixels at the beginning are encoded. Or, in other words, encoding is guaranteed to happen in a single pass.
DEAREST UBQ323, IS THIS A PROPERTY YOU WOULD LIKE TO SEE IN A VIDEO CODEC?
please log in to reply to this thread