Mencoder: Great picture quality, but grainy
 

Hi there,

Tonight I was watching my first KDVD done directly with packshot and no script at all on my TV set (not my PC). And I'm a little disapointed.

I did "We were soldiers" (btw I recommand you to tests any script or other things on War movies like this). The PAL DVD is top quality.

The result with mencoder is really great even if I work on 1/2 a DVD (~2.2 GB). I used max=4500 for that.

The problem is : the picture is REALLY grainy ! Trully like if sand has be sprayed on all the screen. It's boring especially during night scenes where the sky seems to be full of buterflies and other flying bugs.

I don't know what to do. I didn't do any resize so lanczos is not to blame. I used "Noise reduction" set to the second tick and Blockbuster to the first one.

Any suggestion ?

Re: Great picture... but grainy
 

Hi Phil,

I was going to ask you just that. You say you had the Blockbuster setting on the first tick, that is, all the way to the left, right :?:
If you did, that's the minimal noise setting. So I assume the noise you are seeing is not from the noise generator :!:

-kwag

Re: Great picture... but grainy
 

No, let say that for me the initial position is 0.
I set NoiseR to 2 and Blockbuster to 1.

Wich noise generator ? There is an other one besides the blockbuster one ?
Oh do you mean I forget to remove the noise generator I setted for ffvfw a long time ago and is still active ?

That can be that :-(

Re: Great picture... but grainy
 

No. Blockbuster (In PackShot) is the "noise" generator parameter for mencoder.Seems you did use the correct tick values. That's exactly what I use, and I do see (but barely) the noise.
Could you cut a small clip of your encode :?:

-kwag

Re: Great picture... but grainy
 

Wait :!:
Is that the resolution you encoded :?:
Half D-1 :?:
The "Noise" value I tested, was at 704x480.
Probably at 352x it has to be reduced :idea:

-kwag

I think he means by saying "1/2 DVD" that his streamsize is half dvd, therefor not resolution?? As I know Phils workouts I think.

Well Phil very strange, but I shurely do assume that this "grain" is also to be seen on a preview in WinDVD or whatever used on a PC screen??

I also never had that before.

The values of the 1st tick of the blockbuster/noisegenerator trackbar are the one suggested by Kwag, and these are that small that they will be quantized out, well same purpose as blockbuster in avisynth you do use also ... hummmm.

Could you encode a "sampler()" based 2% movie sample but then switching all NR and Blockb. off ... just for comarison?

Re: Great picture... but grainy
 

No, I used 704*576, picture is 2.35 anamorphic (so blackborders = 72 pixel each)

I changed my ISP 2 days ago and I will try to put a small sample on the 50MB space I have now. Let see this we.

@Inc
The grain is viewed also on the PC but I never trust what I see on the monitor because I know the result differs a lot on the TV.

I'll try to put some snapshot too.

Okay, No snapshot needed, I just stop to encode this source.

In fact I did several attempts today and the best picture I obtained was with no noise added and a post processing activated (noise reduction still on the second tick).

But the result was still grainy. So I returned to the source... And against all waiting... all comes from it !

You can't notice the grain at first (and I didn't even noticed it when I watched the DVD on my TV) but whith a little gama level enhance, you have it ! All grainy ! And the phenomena is greatly magnified by the encoding !

So I'm doing a last attempt with noise reduction at its maximum under packshot. I'll see how it's come out.

Note:Do you have also a result that is far below the asked avg ?
First attempt : askef 1930, obtained 1800. Second attempt : asked 1740, obtained 1590 :-(

1pass or 2pass?? (maybe you already mentioned it)

and ... did you change the lmin value packshot generates it its commandline?

@all

I did some tests yesterday and today on full encodings .... and I dropped TOTALLY 1pass when using Mencoder. 2pass is excellent! But for 1pass (= no vpass=x setted) Im shure the parameters we did find out are not optimal as the result is NOT constant quality! Thats why we get very often near exact avg results even when doing NO prediciton, depends on a correct settet lmin value. As I explained that once in another thread that Mencoder with these parameters at 1pass varies the bitrate/Q depending to match final size and not to keep a constant quality -as Phil also found that out when one of his encodings at the end became very worse.

2pass for sure. And I don't modify anything but the crop/expand part.

I saw that also with "Blade 2". The first scene is in a night club, with a lot of flashing blue lights and blood pooring by fire devices. A nightmare for the encoder ! The first pass just gave big squares on this but after the second pass, all was perfect.

Tp clos ethis toppic : I finally did the movie with maximum Noise reduction under packshot (hq3dnr=10:8:2 I think) and no added noise. Things are correct. Not great but very nice according to the average bitrate wanted (1800 for a 704*576 frame res).

I think that will be a little difficulty for people to use mencoder : as we don't use any script, the settings must be tweaked for each source much more than we used to. When you use a script, the quality of all sources is standardized and only really bad sources claim for a tweak in the script.

Well and yours (this one) would be a bad source also for avisynth :wink:
As in a such grainy source state on little moving scenes which would "almost" switch off tempsoften in MA (even if it runs now continously) ... that grain also still would exist in your encoding (IMHO) .... but maybe you did test that source already on MA!?

I think your source is one of the really tricky ones ... (big amount of noise and NOT directly noticable).

No, it is not. Deen removes it very well.
In fact, after gama boost, I saw taht this noise is nothing else than somethign similar to what we add with blockbuster to avoid DCT blocks ! It's possible that a lot of commercial DVD uses this and we never notice it before.

I tested the MA part only of the script in a attemp with CCE. The result was better on this problem but CCE could'nt do what I wanted : my sample was 21 MB long with mencoder and with Q=40 it was 34 MB long in CCE :!:

I hope too ;-).

You've probably allready tried it but why don't you use your avisynth script with mencoder?

-Koekies

Because the purpose is not to use a script but the internal filters of mencoder.