Log in

View Full Version : CCESP and Variable Quantisation - have we REALLY got the CCESP method sussed?


Arky
16th June 2002, 03:49
TactX recently posted this quote:

[Quoted from the Berkeley Multimedia Research Center MPEG FAQ

Non-linear macroblock quantization was introduced in MPEG-2 to increase the precision of quantization at high bit rates (hence, low quantiser_scale values), while increasing the dynamic range for low bit rate use where larger step size is needed. The quantization_scale_code is switchable between the linear (MPEG-1 style) or non-linear scale on a picture coding (frame or field) basis. This new MPEG-2 non-linear scale corresponds to a dynamic range of 0.5 to 54 with respect to the old linear (MPEG-1 style) range of 1 to 31.

In other words: for MPEG-2 use non-linear.]






...and I myself posted this excerpt, from the Spruce MPX3000 manual:

["There are two types of quantization available in MPEG-2 video encoding: linear and non-linear (which is the default). MPEG-1 useslinear quantization only. MPEG-2 has both types of quantization available - though non-linear is used almost exclusively, as it provides greater dynamic range.

Note: When MPEG-2 was defined, the non-linear quantization scale was introduced as an across-the-board improvement on the previously used linear scale. The option of linear scale in MPEG-2 was defined only to accommodate full backward compatibility with MPEG-1. **There are no conditions known under which a linear quantization type would be preferable to Non-Linear type in MPEG-2"

**********************************************************************
I do not offer this contribution as a criticism of Robshot et al's method - quite to the contrary, I congratulate them on a job well-done - a decent guide to cinemacraft, for us enthusiasts, has been needed for a long time and i was thrilled to see the excellent Robshot et al guide. It is the very nature of our little community that we advance our hobby through discussion and experimentation, and it is from this standpoint that I would be eager to see the Robshot et al Cinemacraft method/guide developed/honed further. Make of it what you will. ]







SO.... I am wondering if the last few months have brought us any closer to a method encompassing the thorny issue of variable quantisation. It seems a great pity not to capitalise upon an area of the MPEG2 spec which was clearly devised for a good reason, and which CCESP supports.

Comments please!


Arky ;o)

StephLG
19th June 2002, 10:43
Hi,

I may be a little bit out of topic but I'd like to give an advice to users who encounter an overflow error on a particular frame while trying to encode with CCE (2.62 or 2.64), as Ekin described in a this recent post http://forum.doom9.org/showthread.php?s=&threadid=27287 .

I also encountered this error in CCE 2.62; after a successfull 1st VBR pass, the 3 passes processing always failed after a few seconds with an overflow error. I had followed the 'Robshot method' and started being confused. Then I read in CCE userguide that 'Linear quantization' was meant to be used with MPEG1 encoding, whereas MPEG2 should use 'Non-Linear'. So I unchecked the 'Linear' option and the 3 pass processing has completed without any error. I had tried many other things before that and nothing worked. I really think this is the 'Linear' option which, in some circumstances, gives this error.

For your information, my source file was a 45Gb Huffyuv avi (with audio, even if I don't encode audio in CCE), 720x480@23.978fps NTSC; I made the 1st pass with 8000b/s as max; 3 other passes with min=1500,avg=6000, max=8000.

Sorry Arky if I'm out of topic; feel free to do what you want with this post.