Log in

View Full Version : Finding the optimal bias automatically in DoCCE4U ?


shiva-x
19th May 2003, 23:11
I dunno exactly what the "Auto determine bias setting" option in DOCCE4U does. Maybe I don't use DOCCE4U the way it was meant to, as I only do movie-only DVDs. Anyway here is an example of how I do set the bias in CCE.

Movie : Training Day R2
Running Time : 1h 57m 04s
Calculated Bitrate : 4300 kbps

First pass : ONE pass VBR :
Quantization Factor : Q40.
Min bitrate : 0.
Max bitrate : 9000.

Second pass : Multipass VBR :
1 pass
Click "Advanced..."

Global bitrate setings :
avg : 4300
min : 0
max : 9000

now let's set the bias
at first I set the bias to (AverageBitRate/100), so here it is : 4300/100 = 43.

I watch both avg and max Q factors : avg=3.35 max=3.80
I try to decrease bitrate by one unit (bias=42) : avg=3.34 max=3.86 -> I was wrong trying to decrease the bias.
I try to increase the bitrate by one unit (bias=44) : avg=3.35 max=3.73 -> great :-).

While at least one of the Q factors decreases, the other staying constant (or growing up but very little as when I set the bias from 45 to 46, the avg Q factor goes from 3.35 to 3.36 but the max Q factor goes down from 3.66 to 3.60) or both Q factors decrease, I continue increasing the bias. In this film, I had to decrease the bias until bias=50, where the Q factors are equal and minimal (avg=max=3.34).

So 50 is my optimal bias.

Imagine my bias entry value is too high, my Q factors will either gow down or stay constant when I try to decrease the bias. If both Q factors stay constant or get lower, I continue decreasing the bias until they get bigger. I had this problem encoding "Panic Room". My initial guess was way too high. My Q factors were minimal so I had to lower the bias until one of the Q factors grew up.

Using this method always gives me excellent quality results. Of course, someone can say this is because I use a very high bitrate for encoding, which is not totally wrong. Anyway I can spot the image defects when I set the bias too high (distorted images) or too low ("mosquito" noise).

I posted this thread because I think it could be a great feature to add in DOCCE4U. I think this method of smart bias setting could be automated (Or maybe DOCE4U uses this method already and I lost 2 hours writing this unuseful text :confused: ).

BBWoof
20th May 2003, 18:32
Originally posted by shiva-x
I dunno exactly what the "Auto determine bias setting" option in DOCCE4U does

Auto bias in DoCCE4U is set based on a formula created by Aquabubble which is in turn based on some graphs on Trilight's website.

BBWoof

shiva-x
20th May 2003, 21:57
Originally posted by BBWoof
Auto bias in DoCCE4U is set based on a formula created by Aquabubble which is in turn based on some graphs on Trilight's website.

BBWoof

My method is inspired from those graphs too. But I was wondering if the optimal bias setting was always the same, if it couldn't be different depending on the movie. That's why I used this method as CCE calculates the Q factors in advanced settings so I can tweak my bias manually.