Frame Type (number)

[AQUAR]

Very short question.
What is the significance/meaning of the number inside the brackets behind the Frame type?

If its a long answer - happy to be just pointed to some source material that refers to this parameter.

[Jan Gruuthuse]

Update: see answer by ajschult: http://avidemux.org/smuf/index.php/topic,16185.msg70495.html#msg70495
kind of frame and number represents the location in current video array.

http://en.wikipedia.org/wiki/Inter_frame

All of these frames look complete to the user. But the frames often can not stand by themselves. They usually need information from other frames in order to make their presentation complete.

source: http://wiki.multimedia.cx/index.php?title=Frame_Types

More or less my understanding as user.
Also explains why you can only cut on certain frames and not on others, as key information for prediction of frames would not be available in the cut video.

However clips I tested so far only have either always showing (1) in clip or (2) in clip, frames used for prediction? There is no variation found of number in a clip on this side.

[AQUAR]

Thanks Jan,

The material in those references, and the Avidemux Wiki, do a fine job of describing the various frame types and their interactions.
I do realise that the data parcels for each frame is not stored in the same order as per their presentation order, with AVC being the most complex.
Hence the 2.6 branch giving up this frame ordering control to the decoder.
 
But, I still can't find a correlation to those numbers in the brackets.
For ease - lets call this numeric information the Frame Type Number ie "FTN".

Some media files have frame type: xx (03) and I've even seen a higher FTN.
Stepping through a GOP of an AVC video stream, the FTP stays almost always the same across the different frame types (I,P &B).
But every so often the FTN changes for a single frame.

I note that the majority of my media files just show an FTN of (00) across the board.

I don't think the FTN correlates to the number of frames used in completing/predicting a P or B frame.
If it did there would be more variation in the FTN between the frame types in the GOP.

It also doesn't correlate to the number of reference frames allowed to be used by the B and P frames.

A frame type: I (02) makes no sense in this regard either, as it should stand without referring to other frames, especially Idr frames.
Although I've read that I frame can also be segmented across a few frames, for beter inter frame decoding, in transport streams.

Does this TFN correlate to something at a deeper technical level than that covered at the "GOP" level?
Is that what you are indicationg with "location in video array"? (I need more spoon feeding or missing something obvious).

[ajschult]

The number that gets printed is the _Qp field from ADMImage, which says (avidemux/qt4/ADM_userInterfaces/ADM_gui/ADM_image.h)

Code Select Expand


uint32_t            _Qp;                /// Average quantizer for this image, Default=2


[AQUAR]

Didn't think of snooping at the source code.
Not that I know all that much of C++.

Now I know it refers to the scaling matrix used I can sleep easy. 

Thanks for the response.

[Jan Gruuthuse]

avidemux wiki on average quantizer: Guide to encoding types

[AQUAR]

The avidemux wiki also has an H.264 encoding guide.
http://www.avidemux.org/admWiki/doku.php?id=tutorial:h.264
That guide covers a bit more detail on average quantizer encoding under "Single Pass - Constant Rate Factor:".