[theora] Fixed Quantizer - Fixed Quality

Christoph Lampert chl at math.uni-bonn.de
Tue Mar 25 13:51:39 PST 2003


Hi,

I guess this thread is too long already, this is my final posting to
this. Just my 2 cents, on what I consider "true.

I'm sure that 

-------------------------------------------------------------
1) Fixed Quantizer != Fixed Quality. 

I attached two almost random snapshots from VQEG sequences to show what I
mean: Have a look at barcelona-q20.jpg and suzie-q20.jpg
Both are the first frames of clips encoded with DivX at fixed Quant
of 20. And, no, a quant of 20 is _not_ too high, We'll see why: 

a) "Barcelona": Colorful, many small "randomly" moving object. 
Image quality: surely isn't pretty, but watchable. 
Quantizer 20 is very well possible for this clip: At fixed quant 20 
is has a bitrate of 493,9 kbps(!) at 352x288 at 25 fps (0.194 bpp)

b) "Suzie": low motion, large areas of smooth texture.
Image quality: absolutely horrible. 
At fixed quant 20 it's 98.6 kbps at 352x240 29,97 fps (0.038 bpp)

Of course, when encoding Suzie alone, Quantizer 20 might not happen,
because there are enough bits to spend for a higher quant. But in a full
movie, scenes line a) and b) might be mixed. Fixed quant will not be
optimal. 

---------------------------------------------------------------

2) Encoding with rate control as in single pass "Bitrate control" will not
lead to better quality than fixed quant (with the right value of the fixed
quant). Ratecontrol doens't know anything about "quality". It will try to 
reach more-or-less CBR. 

But somehow this is not a fair comparison, because how do you determine 
the right quantizer value? You have to look at the material, so you have 
extra information. 

---------------------------------------------------------------

3) Two-pass-encoding with varying quantizer can lead to better overall
quality than fixed quantizer encoding. 

E.g.: Encode Barcelona with Quant 25, but Suzie with quant 8. 
Total size will be similar: 

Suzie-Q8:  275442   +   Barcelona-Q25: 347980    =   Total 623422
Suzie-Q20: 115378   +   Barcelona-Q20: 550760    =   Total 666138

But visual quality makes a real difference as you can see from th other
attached pictures: Barcelona-Q25 isn't too much worse than Q20.
Suzie-Q8 is _much_ better than Q20. 

These are just examples, of course...

Christoph 

<p><p>
<hr noshade>
<img src="att-0032/01-barcelona-q20.jpg" alt="barcelona-q20.jpg">
-------------- next part --------------
A non-text attachment was scrubbed...
Name: barcelona-q20.jpg
Type: application/octet-stream
Size: 14972 bytes
Desc: not available
Url : http://lists.xiph.org/pipermail/theora/attachments/20030325/259377dd/barcelona-q20-0001.obj
-------------- next part --------------
A non-text attachment was scrubbed...
Name: suzie-q20.jpg
Type: application/octet-stream
Size: 3583 bytes
Desc: not available
Url : http://lists.xiph.org/pipermail/theora/attachments/20030325/259377dd/suzie-q20-0001.obj
-------------- next part --------------
A non-text attachment was scrubbed...
Name: suzie-q08.jpg
Type: application/octet-stream
Size: 6827 bytes
Desc: not available
Url : http://lists.xiph.org/pipermail/theora/attachments/20030325/259377dd/suzie-q08-0001.obj
-------------- next part --------------
A non-text attachment was scrubbed...
Name: barcelona-q25.jpg
Type: application/octet-stream
Size: 11506 bytes
Desc: not available
Url : http://lists.xiph.org/pipermail/theora/attachments/20030325/259377dd/barcelona-q25-0001.obj


More information about the Theora mailing list