On Tue, 9 Jun 2015 17:11:29 +0200
Carsten Kurz via DCPomatic <dcpomatic(a)carlh.net> wrote:
Am 09.06.2015 um 15:53 schrieb Carl Hetherington via
DCPomatic:
The easiest way to investigate would probably be
to compare a JPEG2000
file encoded by OpenJPEG with a Kakadu one... can anyone provide those?
Do you mean J2C files or MXFs? If I get it right, I will only get J2Cs from DCP-o-matic
when the 'push-to-disc' occurs. Yes, I could unwrap them...
Here are two smallish 4k DCPs with the same pattern, one encoded with DCP-o-matic, one
with Kakadu/FinalDCP.
https://dl.dropboxusercontent.com/u/5582554/4kCheckerFlatOpenJPEG.zip
https://dl.dropboxusercontent.com/u/5582554/4kCheckerFlatKakadu.zip
I'm wondering how IMBs/J2C decoders actually select the resolution layers. If they
simply have a numbering scheme instead of a formal 2k/4k qualifier, how would the IMB
decide wether 'layer 6' is actually 4k or 2k? The Sony Projectors are all 4k, so
for them it would make sense to simply choose the highest number. A DLP system would need
to choose layer 6 or 7, but what if a layer 6 encoded with Kakadu is 4k, and with openJPEG
it is 2k? The encoding parameters for Kakadu and OpenJPEG seem to make that difference
between 6 and 7...
NeoDCP player has no trouble to find/decode the 4k layer from both OpenJPEG and Kakadu.
This is no formal evidence, but I think the bug is in the Doremi IMB software.
Thanks for these. The Kakadu sample has an extra resolution layer (7
instead of OpenJPEG's 6) according to j2k_dump. I'll have a look in
more detail at some point.
Regards,
Carl