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. |
17th June 2006, 02:11 | #1 | Link |
Registered User
Join Date: Mar 2005
Posts: 155
|
Preserving grain
Hello,
I am using megui to encode mpeg2 to x264, but noticed a considerable quality loss due to deblocking which i put on -3,-2. i know i can disable deblocking but that ofcourse will introduce artefacts. the problem is that in removing grain it removes detail as well, like original: x264: is there any way to preserve grain? this happens even with good bitrates, 1300kb/s at 704x400. i use sharktooth HQ slowest profile except i changed the deblocking to -3, -2. i searched and saw mentioned sharktooths matrix but it didnt help. i am new to megui so please be gentle if there is a setting for this! Last edited by futurex; 17th June 2006 at 02:16. |
17th June 2006, 03:20 | #2 | Link |
Registered User
Join Date: Apr 2002
Location: Germany
Posts: 4,926
|
@futurex that is normal without FGS support you won't be able to preserve this kind of Details @ those Low Bitrates and to turn of deblocking the bitrate in your case is to low too, so you have to life with blocks in high motion sequences or you prefer the detail lose thats up to you another possibility would be to use a higher bitrate.
__________________
all my compares are riddles so please try to decipher them yourselves :) It is about Time Join the Revolution NOW before it is to Late ! http://forum.doom9.org/showthread.php?t=168004 Last edited by CruNcher; 17th June 2006 at 03:24. |
17th June 2006, 03:36 | #3 | Link |
Mr. Sandman
Join Date: Sep 2003
Location: Haddonfield, IL
Posts: 11,768
|
if you can "live" with bitrates around 1600kbps and higher you can try my EQM AVC-HR custom matrix, it should help preserving grain and reducing blocking.
__________________
MPEG-4 ASP Custom Matrices: EQM V1(old), EQM AutoGK Sharpmatrix (aka EQM V2), EQM V3HR (updated 01/10/2004), EQM V3LR, EQM V3ULR (updated 04/02/2005), EQM V3UHR (updated 17/12/2004) and EQM V3EHR (updated 05/10/2004) Info about my ASP matrices. MPEG-4 AVC Custom Matrices: EQM AVC-HR Info about my AVC matrices My x264 builds. Mooo!!! Last edited by Sharktooth; 17th June 2006 at 03:48. |
17th June 2006, 08:49 | #5 | Link |
Registered User
Join Date: Jan 2002
Location: France
Posts: 2,856
|
futurex : why would you try to preseve the "grain" of a so bad looking source ? Seing how overshapped it is, what you see isn't grain, it's plain noise enhanced by the sharpening. If you want noise at playback, just add it with ffdshow. ( Imho, even if you don't want noise, you should add it, because it will enhance the visual impression )
__________________
|
17th June 2006, 09:11 | #6 | Link |
Registered User
Join Date: Feb 2004
Posts: 1,348
|
Manao If you noticied the (very bad) haloing I find it hard to beleive that you wouldn't have noticed the (quite evident) bluring present in the screenshot of x264. It definately doesn't look like noise is all the inloop filter is removing to me, or is mpeg4 avc's inloop filter the first perfect denoiser that never removes details by accident .
|
17th June 2006, 09:46 | #7 | Link |
Registered User
Join Date: Mar 2005
Posts: 155
|
manao: it's not that bad looking and the grain is not as obvious, it's just been zoomed in and then resized for greater visibility :-)
its a digital tv mpeg2 i captured, i'm sure the haloing is also present on the dvd when it gets released ;-) im suprised you didnt notice x264's inloop filter is removing not only grain, but grain from his face, looks like he shaved! i did an encode at 1800kbps, sharktooths matrix, no deblocking and it did preserve the noise for the most part, but i think it'll have to go 2200+ to be comparable to the mpeg2, but really 1200kbps was my limit until i get a bigger hard disk :-( |
17th June 2006, 10:59 | #8 | Link | |
Registered User
Join Date: Feb 2004
Posts: 1,348
|
Quote:
|
|
17th June 2006, 11:08 | #9 | Link |
Registered User
Join Date: Mar 2005
Posts: 155
|
bit of both, but mostly because of detail. then again i'm kind of picky, maybe time to get a new hdd, they're pretty cheap. this doesnt happen with non-grainy sources and i can do about 1300to 1400kbps encodes which are quite close to the source
|
17th June 2006, 11:42 | #10 | Link | |
Registered User
Join Date: Jan 2002
Location: France
Posts: 2,856
|
Futurex : what is the quantizer of the frame you took the screenshot from ?
Quote:
Here's what it would look like with a proper noise added back : As you can see, Lock can recover his weathered tan quite easily with a bit of noise added.
__________________
|
|
17th June 2006, 12:32 | #11 | Link |
Registered User
Join Date: Mar 2005
Posts: 155
|
thanks manao, but i assume this would be done in an avisynth script (pp during playback may be troublesome for the pc) which then gets encoded to x264 so won't it end up with the same blurring effect/detail loss?
the problem is, it's the actual texture that is wiped out/blurred, which i fear cannot be recovered through any avisynth script :-( Last edited by futurex; 17th June 2006 at 12:34. |
17th June 2006, 12:40 | #12 | Link |
Registered User
Join Date: Jan 2002
Location: France
Posts: 2,856
|
Here is the grain lost during the encoding :
And here's the grain I generated ( Addgrain, zoomed ) : As you can see, there's not that much of a texture in the first screenshot, and none in the second.
__________________
|
17th June 2006, 13:58 | #13 | Link |
Registered User
Join Date: Apr 2002
Location: Germany
Posts: 5,391
|
By chance, I've 2 pics of Locke on my USB-stick where I showed something to one of my collegues at work.
First: Screenshot from original mpeg2 source of "Premiere" (Pay-TV). Original is anamorph. This *should* be *supposed* to be the better source of them two. Streamed from "Pro7" (Free-TV). Original is "only" 16:9 letterboxed, with a big "7" logo. This should be supposed to be the worse of them two sources. Screenshot is from encoding to "outdated technology" (XviD) @ 1400 kbps, after some slight filtering. Zooming you may do on your own. (The "strange" AR is because of playback thru TV-out /w overscan compensation ... it's okay like that.)
__________________
- We´re at the beginning of the end of mankind´s childhood - My little flickr gallery. (Yes indeed, I do have hobbies other than digital video!) |
17th June 2006, 14:06 | #14 | Link |
Registered User
Join Date: Feb 2004
Posts: 1,348
|
This is a screenshot of the amplified difference between the source and a divx encoded version of a scene from the matrix revolutions (a virtually noisless source)
Notice how it looks like all that was lost in areas that contained texture was noise, but anyone who has viewed a Q4 divx file can tell that more then "noise" is removed from the picture. |
18th June 2006, 10:53 | #15 | Link |
Registered User
Join Date: Apr 2002
Location: Germany
Posts: 5,391
|
The point I was trying to make is NOT: my XviD encodes look better than all you guys x.264 encodes ...
The point I was trying to make is this: Encoding to a lossy encoder will lose information. Always. No way around that. Therefore, it's a valid strategy to slightly bump up those kinds of information we want to keep (detail), and reduce those kinds of information we don't want to keep (noise et al.) before feeding the source into the encoder. This greatly helps the encoder to produce an output more close to what we would like to have. In particular this holds up for DVB broadcasts, at least for the ones I'm used to get from satellite: grain is distorted by the realtime recompression, detail is damped down and softened, additional artefacts have been introduced that shouldn't be there in the first place, and so on. (DVB in Germany is faaar away from DVD quality ... more close to that of a n00b's first DVD Rip, instead.) For comparison ... in the thread Very good quality at 10000+ kbps the OP asks by what codec to achieve maximum results at 10k bitrates. However, he is also speaking about a "standard PAL broadcast". Now, it seems like he'll only come very close to the point where the broadcast originally was. Never beyond, and in fact still below. However I suspect that, when fondling the source with a little Avisynth, the final experience would be better than that of the original broadcast, instead of "just as good". At 5k bitrate instead of 10k, probably. Sorry for the small excurse. But video encoding is a complex topic ... and if one is using only half of those puzzle pieces that were in the box, the puzzle will keep being holey. P.S. In case anyone is wondering ... what I used for the LOST people, after cleaning-up & generalization has become known as "SeeSaw()".
__________________
- We´re at the beginning of the end of mankind´s childhood - My little flickr gallery. (Yes indeed, I do have hobbies other than digital video!) |
19th June 2006, 06:22 | #17 | Link |
Registered User
Join Date: Apr 2005
Posts: 1,339
|
Here's a short (<10 sec) vid only source you can experiment on:
8.5mb http://rapidshare.de/files/23464355/jl.zip.html |
Thread Tools | Search this Thread |
Display Modes | |
|
|