Alright, I'm logging this as a draft so this is more of a log. This is going
to contain a lot of info.
Before 7:25 - Being a total losser engineering student not in the mood to go out
drinking I decided to walk around my front room in my apartment with my dslr in
continous mode
7:25 Started encoding uncompressed 8 bit video of 902? frames at 24fps with
virtual dub. The source footage was fine compressed medium resolution from my
Nikon D50 dSLR camera. Its essentially my front room as fast as my camera can
shoot and me walking around with a 50mm F1.8 lens at 1/30 a second and ISO 400
Estimated file size.... 8709.2MB... should be that since its uncompressed
Resolution is 2256x1496 pixels, its sRGB
Slow hard drive means over 11 minute write time, stupid laptop hard drive.
7:29 Checking out oneriver codec site for testing purposes to make sure I can
check my results and trying to make sure I know a difference between visually
lossless and lossless.
7:36 Finished encoding - Filename is D50-Uncompressed.avi and file size is
9,132,711,252 bytes
7:40 - Realized that the output file needs to be divisable by 16 in both
directions... damn!!!! start an uncompressed output all over again
7:46 - used resize to 2256 x 1504.... gone to store for ice cream and fruit for
making some sort of blended treat, estimated time to output 13 minutes
8:14 - got back.... gonna watch bsg then deal with this
9:23 - finished watching bsg, encoding was completed, going to test accuracy....
file compression is hella good.... compressed file is 1,155,798,447 bytes,
uncompressed (due to resize {This was while making the exported file not during
encoding}) is 9,181,549,124 bytes so that is about 8:1 and I very much thing
that this is a much more fair comparison
9:28 - turns out it needs to be RAW and not MP4 coded for avisynth to work...
dang, 34 minutes to encode
9:34 - encoding at 0.29fps.... so horrible... maybe I'll make a milk shake
10:05 - Finished my milk shake, it was good, encoding is almost done
10:10 - A quote from jpeg200info.com
What is visually lossless compression?
The term visually lossless is a misnomer. Visually lossless compression is
actually lossy compression at low rates. It means that the compressed file is
"visually indistinguishable from the original", however there is still loss
involved in the compression. Lossless compression is a mathematical guarantee,
visually lossless compression is subjective and means "close enough".
Just what I thought, if pixel per pixel is identical for value everywhere its
true lossless, Visually is just "damn good lookin!" So what I've proven so far
is true, no disputtin, no fightin, pure high compression lossless period as far
as I am concerned and since nobody can show me a single pixel difference out of
3060507648 pixels then I'm going to say that their is no dispute, its lossless.
10:16 its done encoding, but its just being stupid with all the packaging, so
annoying... soon be able to report the raw stream size... don't expect it to
change much, its still gonna be 9:1 and lossless
10:19 final size is 1,155,797,471 bytes.... you know what that means, 9:1 and
I'll give this a fair test seal (If I had one) actually, an unfair one because
the differences in all pictures is rather significant and forces it halfway to
intraframe land imho.
10:23 - encoutering the same stupid error with unable to find directshow
filter... dang
10:28 - heck, i dunno, it works now?? oh well gonna decompress it
10:29 - 14 minutes to write the file back to hard drive... I expect a single
missing frame at the beginning though for no real apparent reason, happened
before.... acurse the slow hard drives, those shall be my bain
10:34 - got my laudry and folding it while waiting for the export....
10:52 - its done, gonna go test for pixel differences now - gonna make sure i
follow the oneriver codec guide
Its not lossless.... crap...
Update: It may be lossless....
Initially the footage was 4:4:4 RGB, but in order to encode it had to be changed to YV12 colourspace.... Thought it would remain 4:4:4 but YV12 is 4:2:0 colour space so I need to redo tests with this colour space comparison