MPEG2 ESPECIAL NOISE TREATMENT – Avisynth. Sorry!

[Aeolis]

IMPORTANT: I AM REALLY SORRY I'VE NOT PAID ATTENTION TO THE HUGE RED LINES BEFORE WE POST HERE. SORRY, SORRY, SORRY, SORRY!

I TALKED ABOUT “THAT” REFERENCE JUST TO GIVE AN EXAMPLE (BAD EXAMPLE I KNOW, SORRY AGAIN!), BUT I HOPE YOU CAN HELP ME WITH MY “NEW” POST WITHOUT “THAT” REFERENCE BECAUSE I ACTUALLY WORK WITH CG AND MOVIES TO PRESENTATIONS. MOVIES THAT I MADE BY MYSELF. AND I’VE THE SAME PROBLEMS WITH MY (REALLY MY) OWN SELF MADE MOVIES. SO HERE WE GO AGAIN. SORRY ONCE MORE!


Hi folks,

Here is a big one!

I’ve a personal CG movie. It’s split into two parts. I’ve encoded it doing the following:

I’ve opened both files with VirtualDub-MPEG2 1.6.11 applied Resize to 720X480 (Lanczos3), then frameserve then as 1.vdr and 2.vdr.

Extracted the two .wav files form the movie part as 1.wav and 2.wav. The audio from the original movie parts is .ac3 5.1.

Opened both .vdr files with VFAPI and generated the two dummy .avi files to feed in VEGAS 6.0.

I’ve fed the two audio files in VEGAS 6.0 too.

I frameserve to VEGAS so I can convert any frame rate to 29.97 progressive without worries and sync the joined audio and video too.

So I frameserved the joined movie from VEGAS to CCE 2.70.02.01 using Debugmode Frameserver and encoded it with VBR MIN/AVG/MAX 7000/8000/9000 kbps.

At last, I’ve encoded the movie audio with VEGAS DOLBY plug-in as .ac3 5.1, 384 kbps.

Finally I’ve authored with DVDLab PRO 1.55.

Size is not a problem I will record it in a DVD+R DL. I’ve just to keep files under DVD-VIDEO Specs.

I’ve recorded a DVD+R DL with all that stuff following RecordNow! 7.3 instructions given by Edmund Blackadder. It worked like a charm for me. Follow it I you want to burn DVD+R DL with Videos search for DVD+R DL burning and you should find it in VIDEOHELP Forums.

It played fine it was under DVD-VIDEO specs.

!!NOW THE PROBLEMS!!

The final movie has perfect pictures moments you can think it is a perfect original one, but especially in dark flat areas (they are frequent in this movie it is a CG Movie) I got little blocks formation.
I’ve already searched for filter here and in google and found very interesting things. I’ve tried MSU Deblocker (old 2.2 and new smart 0.1) they looked fine but two blurry no better results with VirtualDub custom filters. So in google I found out that that noise I have could be generated (not completely) but in most part because of MPEG2 encoding.
Something about flat areas are low-frequency concentration and encoders as CCE try to compress it more then the normal and end generating that noise. The solution told for those guys in google found page is to input a controlled noise in the source so in this way you are increasing the flat areas frequency forcing the encoder to do a better job and reducing the black flat areas block noise. An Avisynth plug-in that have this purpose is blockbuster plug-in. I could not find any to VirtualDub.

So here are my questions?

1. Is there a similar plug-in for VirtualDub?

2. Could somebody post a Avisynth script or better the scripts and instructions to accomplish the task I told in this post first part without doing too many frameservings as I actually do and specially without VEGAS at least for the video part? I think that the audio encoding I will still use VEGAS.

Please help me. Remember they two parts so the instruction and script should that join part and the filtering too including the blockbuster filter settings. And if you have suggestion of pre-filtering the source put it in the script. LET’S WORK TOGETHER AND SOLVE THIS QUESTION! MAY BE THE FORCE BE WITH US!

NOTE: I know what I am asking is a big task, but please don’t humiliate me. At least post the script indications I could test the particular settings.
And remember as far as I know a lot of people have this black flat areas block noise problem and I think this post reply could help all of us especially with the introduction of this Avisynth blockbuster idea found in google.

[Abond]

Oh, my god! If I wanted to do this I would at least change my nickname!

[BJ_M]

depends if the blocks were there before encoding to mpeg2 , prob. not ...

and if there were not - it is your encoder settings

[jimmalenko]

I’ve a personal CG movie. It’s split into two parts.

If you made it yourself, why did you split it into two parts ?

... Oh, and why is this post almost word for word like your other post ?

Any connection there ?

[Aeolis]

Hi folks,

To BJ_M: Thank you for your attention and comprehension. You are right that blocks are from the CCE encoding. Let’s share more ideas here everybody that have good ones.

And to Abond and Jimmalenko: Do you really think I would rewrite everything again? I have other things to do, men. I don’t have anything against you, but could you come here with ideas and not criticisms? I’ve apologized in the beginning of this post and I’ll do it again. I AM REALLY SORRY! And I have - yes - different CG movies that I’ve done myself that I would like to join, but as I said I didn’t have time to rewrite that post again changing all the necessary details, the idea is the same so I did it.

AND ABOUT THE POST: I’ve solved the black flat areas block noise as well the block noise in other areas to. Well, I said I’ve solved, but all the greetings go to google and, of course, to the inventor of Blockbuster Avisynth filter. Man this filter really rocks!!! It solved all the blocks and I haven’t tested all the possible settings yet, I’ve just used the noise method with standard settings. It’s really amazing! Well, I still would like to have people sharing their ideas here so I could share mine too. I have other questions in my post could somebody help?

See you later,

Aeolis

[Abond]

Your procedure involves too many programs and too much frameserving. I would do it with only VDub (VDubMod actualy), Avisynth and CCE. Maybe also FitCD.
1. Join the avis with VDubMod. Demultiplex the audio, and save a silent video.
2. Load the silent avi in FitCD to get correct resizing parameters.
3. Write an avisynth script using the cropping-resizing parameters from FitCD and some filtering (I personally prefer to use hybridfupp function)
4. Load it in CCE and encode.
5. Author with the original audio (it is already ac3 5.1).