Bits/(pixel*frame) for AVC?

[kyle1]

Did anyone test AVC codecs to find minimal coefficient keeping good quality? I believe it should differ from MPEG-4 one (0.19).

[E-Male]

without testign i can tell you that there is no "minimal coefficient keeping good quality" for any codec
it always depends on your source, how sharp/noisy/... it is

[Sharktooth]

but depends also on codec settings you use. also AVC has inloop filter that can render the bits/(pixel*frame) formula useless.

[kyle1]

i was asking 'cos i'am using GK for DVD backup. Quality of content is more or less constant there. Well, thanks anyway.
AVC has inloop filter that can render the bits/(pixel*frame) formula useless
Can you suggest more usefull formula?

[E-Male]

...AVC has inloop filter that can render the bits/(pixel*frame) formula useless.
can you go into some more detail on this please

[yaz]

ehrm ... some things should be kept in mind when considering bpp. yeah, it's just history but worth to know. the 'concept of bpp' was developed in the heroic age of divx3.11 sbc (~nandub). that time xvid was just 'walking in kindershoes' (devapi2, maybe) and divx team (called project mayo or sg like that) was just releasing predecessors of divx4.

cutting it short, the bpp findig was 'if bpp is set up in a certain range it will produce reasonable compression (~>70%) in a 2pass encoding' ... very likely(!) ... but if not, the set up must be tweaked further. so, the final goal was maintaining reasonable compression w/the least trials and bpp gave a good starting point to that. all this was based on hundreds(!) of experiments and not on any theoreticals (neither on plain speculations). the compression limit based on the general agreement that quality below this range is unacceptable.

... but things have changed (as usual)
- the codecs these days are much more powerful, armed w/lots of sophisticated options (like qpel, vhq, multiple bframes, reliable rdo, aq, ...) making possible to lower the acceptable compression limit. the devapi3 branch of xvid was able to produce pretty nice outputs round 60% or sometimes even below.
- a series of cqms were developed. selecting the appropriate cqm compression can be lowered much more while keeping acceptable quality.
- series of tools (filters) were developed for preprocessing the source in much finer ways than that times.
- ...

anyway, the concept of bpp is still alive, and as the examples show (say, agk) it can be applied for mpeg4asp w/succes. but adapting it to avc would need much more experiments than for divx3 as there are much more options to be considered.

however, a good strategy (like bpp or sg quite different) for 'how to set up avc encodings for getting good quality' would be warmly welcomed from my side. i know lots of 'guides' about it but none of them convinced me.

the bests
y