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. |
|
4th April 2009, 18:11 | #1 | Link |
Registered User
Join Date: Mar 2009
Posts: 4
|
Pls help me understand B, I & P frames and psy-rd
DISCLAIMER: I'm a Newbie; please bear with me...
1. I've read and read and read, but I still don't understand B frames, I frames or P frames. One site advised using as many B frames as possible because they're inherently good somehow (how, IDK). Another site said not to use too many B frames because they result in a less sharp picture. What are the differences between these frame types, and what is a good number of B frames to use when decoding with, say, HandBrake? (That doesn't result in a less-sharp picture.) 2. What is "psy-rd" and what is a good setting? I hear the settings are very fussy and you can get a fuzzy, less sharp picture unless you use psy-rd...just...right... 3. Does the very latest H.264 support psy-rdo, and what the heck is it? What are good settings? 4. Reference frames: the more the better? 5. Does anyone know the H.264 instruction for "verbose logging"? I've tried v=1, vb=1, verbose=1, verboselog=1, verboselogging=1 in HandBrake, but the Activity Window always says unrecognized instruction -- every time. Sorry for so many questions. Any help is appreciated. Thanks, Kerry |
5th April 2009, 01:50 | #2 | Link |
Registered User
Join Date: May 2006
Posts: 957
|
1) An I-frame (or key frame) is one that does not depend on any other. A P-frame is one that involves prediction/compensation from other previous frames. A B-frame is one that that involves prediction/compensation from previous and future frames. B-frames are good because they are small. With a good encoder/format B-frames don't have to look bad.
2) Psy-rd is a psycho-visual adjustment in x264. It try to retain a similar frame complexity as in the original because this looks good. 3) x264 supports psy-rd. What is it? You just asked and I just answered. For more see: http://forum.doom9.org/showthread.php?t=138293 4) Yes but more makes x264 slower and there are level-related restrictions. 5) x264's verbose switch is --verbose (-v for short). In future check x264 --longhelp Finally: H.264 is a video format, x264 is an encoder for that format.
__________________
x264 log explained || x264 deblocking how-to preset -> tune -> user set options -> fast first pass -> profile -> level Doom10 - Of course it's better, it's one more. |
5th April 2009, 05:19 | #3 | Link |
Registered User
Join Date: Mar 2009
Posts: 4
|
psy-rdO
Thanx, J_Darnley.
Your answers were VERY helpful. Question #3 was about psy-rdo. Is it the same as psy-rd? If not, what is it? Do the latest x264 encoders support it, and what are ideal settings? Thanks again, -Kerry |
5th April 2009, 18:13 | #4 | Link |
Registered User
Join Date: Mar 2009
Posts: 4
|
J_Darnley,
Thanks for your help. You said that with a good x264 encoder, B-frames don't have to look that bad. I use 2-pass encoding and set the "Average bitrate (kbps)." If I increased the Average Bitrate, would B-Frames look better (sharper) or would it have no difference on PSNR? I am presently encoding episodes of The Simpsons (Seasons I BOUGHT and PAID FOR) to create a "Best Of" DVD containing only my favorite episodes. That will save a lot of searching through discs and seasons and ejecting and re-inserting discs to get to my favorite episodes. The Simpsons is not the most complex animation in the world--lots of solid colors. I am using "Reference Frames" = 9 and "B-Frames" = 8. (Plus "Mixed References," "Direct Prediction" = Automatic, "Weighted B-Frames," M.E.M.=umh, "Motion Estimation Range" = 64 (the max), "Subpixel Motion Estimation" = 9 (the max), "Analysis" = All, "8x8 DCT," "Trellis"=2, "No Fast-P-Skip," "No DCT-Decimate," and "CABAC Entropy Coding." Is my Reference Frame value too high? And is my B-Frames value too high? Thanks Last edited by Kerrin Shea; 5th April 2009 at 18:17. |
5th April 2009, 18:00 | #6 | Link |
Registered User
Join Date: Mar 2009
Posts: 4
|
Thanx, Merlin777.
That clears the issue right up! I'm still learning, and I appreciate the help of experts such as yourself and J_Darnley. My philosophy is, I could do lots and lots of "leg work" and find the answer myself, but going straight to an expert for the answer saves me time, effort and "leg work." Experts are like "aliases" or "shortcut" icons. They take you directly where you want to go with no hassles in between. Thanx, Kerry |
5th April 2009, 20:37 | #7 | Link |
Registered User
Join Date: Dec 2005
Posts: 1,460
|
More reference frames usually help on animation but for normal content more then 5 is probably not very helpful. Generally from each reference frame you add you can expect about half as much quality gain then from the previous. In my opinion it's mostly the question how much time you are willing to spend on encoding and if you need hardware compatibility, in which case you have to choose an ammount appropriate for the resolution and level you encode to. Since you encode cartoons there are some things you should do, like lowering the aq strenght. Use search, it was discussed recently.
Also, drop "No Fast-P-Skip" and "No DCT-Decimate", they don't really help. Trellis 1 will also be much faster compared to the quality you lose. |
5th April 2009, 21:08 | #8 | Link | |||
Fighting spam with a fish
Join Date: Sep 2005
Posts: 2,699
|
Quote:
Quote:
Quote:
If I may ask, what are you using to encode? MeGUI comes with a plethora of built-in profiles that can tackle anything you throw at it. Each of these profiles were made by our very own Sharktooth, and have been tested repeatedly for errors. |
|||
Tags |
b frames, h.264, psy-rd, psy-rdo, reference frames |
Thread Tools | Search this Thread |
Display Modes | |
|
|