KVCD and the quest for the De-Noiser...
Wanted to keep this out of the preset forum.
I haven't used a denoiser, and never seem to have any problems. However, I will check one out tonight, the one that you suggested, and see if it clears up the picture. Most of my encodes are backups from DVD VOB's, which is a pretty darn clear picture to begin with. But, if you have better luck with any of the filters, give me your opinion, I am always open for better quality. Also, what kind of TV are you watching these on? I have a Mitsubishi 55" HD WS, and a Mitsubishi 27" Tube The picture on the 55" is good, but not as good as DVD with the 480x480, on the 27", I can't tell any difference... I mean no difference what-so-ever. |
Re: Attn: Dialhot and the quest for the De-Noiser
Quote:
More over : the noise is introduced BY THE DECODER. Noise if the result of rouding into mathematical operation performed during the render of the picture. The error diffusion that is introduced is embed in the MPEG2 specifications and can't be avoided. That is why you NEED a denoiser. I suggested you undot but removegrain(mode=2) is better. And you should try also deen. Personally I use both for all my DVD->KDVD conversion ! You are doing KVCD so, in theory you need more filters that I need ! Quote:
|
I'd go for RemoveGrain(mode=2).TemporalSoften(2,3,3,15,2) or RemoveGrain(mode=5,modeU=2).TemporalSoften(2,3,3,1 5,2). I've found Deen to be a bit aggressive in some cases, and I prefer temporal-only denoisers :wink:
|
I've substituted in my last 8 KVCDs Undot with RemoveGrain(mode=2), or (mode=8, modeU=2), with speed like Undot and more compressibility. The rest of filters are those in the MA script.
I finished yesterday my first KDVD with RemoveGrain(mode=2) in it. I robot and Troya. Trpya looks great. But I robot... I see frequently a kind of flickering, like two repeated frames, I cannot explain. I reencoded with undot, and now doesn't happen. I saw this effect reported with RemoveDirt. This is the first time I saw this with RemoveGrain (I think this is the guilty one, since the effect was corrected after substituting it). Has anyone seen this effect?, do you know a way to avoid it?, I would like to keep with RemoveGrain... In my script I also used BicubicResize, and ScriptClip...Unfilter...TemporalSoften. |
You could try mode=1, it's supposed to be exactly the same as UnDot with one small bug corrected. Mode=5 is between 1 and 2 so it might work as well. There probably are thin lines in the frames which are mistaken for noise/grain/dirt.
|
Hello,
just another denoiser question, lately ive been using Dust() for my captures, only one problem,its damned slow.(But the result is damned good!). about 23 hours encoding time for a one hour capture. heres the script: Quote:
the denoising job just as good as Dust()? Yeah I know testing is the way to find it out,but I like to hear some opinions first :D thanks ---------------- viking |
Take my advice : drop all old filters forever and start using new ones. At first you will be convinced that you can't do the same things with new ones but that's only a mental lock !
Deen, RemoveGrain (not RemoveDirt that has problems), Convolution3D or even Vaguedenoiser (slow but not as slow as Dust !) can do exactly the same things a lot faster. All is matter of time to test and combine filters. As I said once, I work 40 hours just to do my v4 script (for avi->kvcd) and v5 is not yet really finished ! |
Many feel that RemoveGrain().RemoveDirt() does a similar job. It has its problems (the thin lines can cause flickering) but IMO it's a very good denoiser combo. Lately I've also been using RemoveGrain(mode=2).TemporalSoften(2,4,5,15,2) for my analog captures with very good results both quality- and compressibility-wise.
|
Ok,then I make some tests with
RemoveGrain(mode=2). TemporalSoften(2,4,5,15,2) and later other combinations of the mentioned filters. And I guess that TemporalSoften() works fine even if target is KDVD(720(704)x576)? ------------------------- viking |
Yep, my targets are always 704x576.
|
So I did some tests:
1. Quote:
Quote:
min.bitrate 300,max bitrate 5000,higest quality(very slow) setting. It was hard to tell any difference between the two clips, maybe maybe slightly sharper picture using Faerydust(3). But the encoding time with no 2 was only 1/4 of the time no 1 used so I guess ill use no 2 from here. :D The encoded file size was 0.8MB larger with script No 2. But a question;what setting should I use under Video Arrange Method in TMPGEnc?? So far I have used "Full Screen" and it seems to work fine. I read somewhere here that "Center" is used. Anyone of you that can tell me whats the "correct" setting?? thanks -------------- viking |
And you'll encode even faster if you use "High quality" in TMPGEnc. Some tests were made a long time ago and in fact it produces better quality than the slowest method.
The arrange method depends on whether you add the borders in your script or let TMPGEnc do it. I've always used Full screen (keep aspect ratio). |
Quote:
@boulder Personnaly I always use "motion estiamte". Even faster and to be honnest, I didn't have any problem with it (the CQ is a little lower than in High but... I really have a slow PC). But this time I drop quality for encoding time. I'm ashamed about myslef :-) |
Ok,I add borders in the script,
so then it will be ok with "Full Screen".or "Full Screen"(Keep Aspect Ratio),but I guess since we deliver the video with given parameters from a script to TMPGEnc,everything is "locked",like the A/R,borders,resolution etc., so I dont know if TMPGEnc can "destroy" the A/R if I use only "Full Screen" --------------------- viking EDIT: I saw Dialhot's reply after posting this |
Quote:
|
Wow, ask a simple question get a complex answer! I love this forum.
|
Site design, images and content © 2002-2024 The Digital FAQ, www.digitalFAQ.com
Forum Software by vBulletin · Copyright © 2024 Jelsoft Enterprises Ltd.