h.264 level_prefix

malikcis · 25th November 2010, 10:33

Hello,

In Iain E. Richardson's h.264 book "the h.264 advanced video compression standard",
details are missing in "Chap 7.4.2 -> 3.Encoding the levels of the remaining non-zero coefficients"
It states that Level VLC is composed of level_prefix and level_suffix, but he doesn't explain how level_prefix
is derived from level from encoding point of view. This part is hard to be reverse engineered from decoder spec.
Where can I find more precisions regarding this?

This topic has been discussed here previously without 100% accuracy:
http://forum.doom9.org/archive/index.php/t-111926.html

Thanks, malikcis

imcold · 25th November 2010, 16:36

Look for the jvt document called "JVT-D034.doc", it contains example code for level_prefix/suffix construction.

malikcis · 26th November 2010, 16:53

Thanks,
I downloaded the doc.
This kind of doc gives me headache at friday afternoon.
Found a nice PDF on that topic:
http://iris.ee.iisc.ernet.in/Courses...LC_Example.pdf

Cheers

Shevach · 30th November 2010, 14:52

Perhaps, it is interesting to know also what's maximal value of the level_prefix (since the maximal code size of level is derived from level_prefix).

In the document JVT-X210, in the item #83 is said:

Gary Sullivan suggests: People keep asking what the range of level_prefix is for FRExt profiles. It would be nice to figure out the answer and supply it in a NOTE. The same idea applies to other aspects where some helpful (but not strictly necessary) information has been left out of the standard – e.g., ranges of various things that are specified only indirectly but that need to be understood by implementers. Lowell Winger reports on 17 Oct 2006 that the maximum value of level_prefix is =19+(bitDepth – 8).

25th November 2010, 10:33	#1 \| Link
malikcis Registered User Join Date: Sep 2007 Posts: 50	h.264 level_prefix Hello, In Iain E. Richardson's h.264 book "the h.264 advanced video compression standard", details are missing in "Chap 7.4.2 -> 3.Encoding the levels of the remaining non-zero coefficients" It states that Level VLC is composed of level_prefix and level_suffix, but he doesn't explain how level_prefix is derived from level from encoding point of view. This part is hard to be reverse engineered from decoder spec. Where can I find more precisions regarding this? This topic has been discussed here previously without 100% accuracy: http://forum.doom9.org/archive/index.php/t-111926.html Thanks, malikcis

25th November 2010, 16:36	#2 \| Link
imcold pencil artist Join Date: Jan 2006 Location: Slovakia Posts: 201	Look for the jvt document called "JVT-D034.doc", it contains example code for level_prefix/suffix construction. __________________ fevh264 - open-source baseline h.264 encoder

26th November 2010, 16:53	#3 \| Link
malikcis Registered User Join Date: Sep 2007 Posts: 50	Thanks, I downloaded the doc. This kind of doc gives me headache at friday afternoon. Found a nice PDF on that topic: http://iris.ee.iisc.ernet.in/Courses...LC_Example.pdf Cheers

30th November 2010, 14:52	#4 \| Link
Shevach Video compressionist Join Date: Jun 2009 Location: Israel Posts: 126	Perhaps, it is interesting to know also what's maximal value of the level_prefix (since the maximal code size of level is derived from level_prefix). In the document JVT-X210, in the item #83 is said: Gary Sullivan suggests: People keep asking what the range of level_prefix is for FRExt profiles. It would be nice to figure out the answer and supply it in a NOTE. The same idea applies to other aspects where some helpful (but not strictly necessary) information has been left out of the standard – e.g., ranges of various things that are specified only indirectly but that need to be understood by implementers. Lowell Winger reports on 17 Oct 2006 that the maximum value of level_prefix is =19+(bitDepth – 8).