Welcome to Doom9's Forum, THE in-place to be for everyone interested in DVD conversion. Before you start posting please read the forum rules. By posting to this forum you agree to abide by the rules. |
|
|
#1 | Link |
|
ffx264/ffhevc author
Join Date: May 2007
Location: /dev/video0
Posts: 2,019
|
trellis question
Hi,
x264 supports 2 trellis settings. trellis=1 and trellis=2. As I understand it, when using the first trellis setting, trellis only kicks in when needed. What I don't get is how does x264 "knows" when to use trellis and when not? Is there some threshold that must be met in something before trellis=1 kicks in? Also, as I've read here, trellis doesn't work with deadzones. But isn't it possible to do something like... 1) use trellis and get the values 2) use deadzone and get the values 3) compare the values of trellis with those from deadzone 4) use the best value (either from trellis or from deadzone)
__________________
The Cyberpunk Database || NTP & NTS Time Server || Mirror Services ffhevc || ffx264 || ffxvid |
|
|
|
|
|
#2 | Link | ||
|
Registered User
Join Date: Dec 2004
Location: Tomsk, Russia
Posts: 366
|
Quote:
trellis = 2 works every time transform is called. Quote:
trellis tries to optimize RD function (and actually reaches the maximum for given mb), while dead-zones is kinda magic - you'll need (and you can) to tune them up for every movie in order to achieve whatever result you want.
|
||
|
|
|
|
|
#3 | Link |
|
x264 developer
Join Date: Sep 2005
Posts: 8,666
|
Trellis always gives better results, PSNR-wise, than deadzone.
However, for strong grain or cases where you want more sharpness retained, low deadzone settings can look better than trellis (if that's what you want, of course). Also note that Trellis 2 has a habit of decimating grain/fine detail. |
|
|
|
|
|
#4 | Link | |
|
x264aholic
Join Date: Jul 2007
Location: New York
Posts: 1,752
|
Quote:
|
|
|
|
|
|
|
#6 | Link |
|
x264 developer
Join Date: Sep 2005
Posts: 8,666
|
That's the point; its optimized for PSNR vs bits, and as a result can decimate fine detail in favor of bitrate savings.
Yeah, but it was pretty crappy IMO. I need a good "reason" to use a lower deadzone; i.e. a way to tell the algorithm what kind of blocks lower deadzone is good for. |
|
|
|
![]() |
| Thread Tools | Search this Thread |
| Display Modes | |
|
|