Welcome Guest ( Log In | Register )


Important

The forums will be closing permanently the weekend of March 15th. Please see the notice in the announcements forum for details.

 
Video Processing Filter Setup
« Next Oldest | Next Newest » Track this topic | Email this topic | Print this topic
Asmordean
Posted: Aug 30 2002, 04:10 PM


Unregistered









I use my PC as a VCR and like to archive some of the captured data.

After many attempts, I have come up with about the best filter setup for capturing cartoons (anime) that I can do. I thought I would share it with the world.

Capture in 720x480@29.97fps (NTSC) using PicVideo MJPEG codec at quality 17. I tend to use the maximum resolution that I can because I can always resize down if I need to. Nvidia Personal Cinema maxes out at 720x480.

After capturing the video I use the following filters in this order:
-Deinterlace (Blend) - Internal filter (Option here, if the source can be processed with inverse telecine I use that instead. However, that tends to be rare)
-2D Cleaner Optimized (0.9) - Defaults (10/ 2x2)
-Temporal Cleaner - Defaults except for threshold which is lowered from 30 to 10.
-Brightness/Contrast (Bright -25%, Cont 125%)
-Area Smoother (0.1) - The setup of this one varies from source to source. Try to set it so that any painted backgrounds are not effected in any cell.
-Unsharp Mask (1.3) - Diam 5 / Str 50 / Thr 0
-resize (Precise Bicubic) - 400x300
-Compress however you prefer. I use Divx 5 at 900kbps for the video and LAME MP3 at 64kbps for the audio

I store my videos at 400x300. The reason being is that it scales nicely to 800x600 which is my video out resolution and is smaller than 640x480.

Using this method I can get near DVD quality from the SVideo output of my cable box. Although it is smaller, it looks better than the original broadcast.

Feel free to add your own setups below for various sources - Action, drama, cartoon, computer animation, etc.
 
  Top
SillKotscha
Posted: Aug 31 2002, 10:01 PM


smart Moderator


Group: Moderators
Posts: 146
Member No.: 6
Joined: 7-July 02



thanks for your input...

here is my capture and filter chain:

- capture-resoltution: 720x540 (FullPAL without TV overscan)
- for movies MJPEG @ q19
- for short video clips (mtv etc.): Huffyuv or MJPEG @ q20
- Deinterlace MAP v1.0 (motion and pixel) by Shaun Faulds - set to defaults
- internal resize: to 512x384 - precise bicubic (a=0.60)

looks very very good and even better seen on tv via tv-out (ATI AIW)

cheers Sill

--------------------
"Have you ever noticed that whenever Microsoft calls something 'Smart', it's definitely a feature you want to disable!"
 
     Top
vondrack
Posted: Sep 10 2002, 08:05 PM


Unregistered









OK, here's my $0.02:

first:
- capturing at 1/4 PAL (384x288), Huffyuv in YUV2

second:
- 2dcleaner 0.9, threshold 10, area 3x5 (x=1, y=2)
- temporal cleaner, default setting, luma enabled

this runs almost real-time (some 22-23 fps) on my P4/2G.

third:
- use TMPGEnc (CQ, maximum quality) to encode to MPEG-2

The resulting file has about 33 MB/min and looks ABSOLUTELY perfect when played back in a small window on the monitor and still VERY good on my 27" TV... Actually, you realize it is not as sharp as the original TV broadcast only if switching back and forth between the MPEG and a live broadcast... and the MPEG has almost no noise at all, while the live TV has quite a lot of it.

I know I could make the resulting file much smaller if sacrificing a bit of the picture quality, but I am not that short of storage space fortunately... frankly, even files about 25% smaller looked almost identical (on the TV) to what I mention above... I was not determined enough to watch the same M.A.S.H. episode over and over... biggrin.gif so I eventually decided to use the best TMPGEnc setting in terms of the quality (making the MPEG almost undistinguishable from the filtered AVI) and postpone fine-tuning the file length until I have more time/determination.

BTW, SillKotscha... from what I have read on various sites, you should use precise bilinear for your resize. Use bicubic when scaling up, but bilinear when scaling down, they say... the picture should be better and the filter will run faster. Have you tried it?
 
  Top
SillKotscha
Posted: Sep 11 2002, 07:12 AM


smart Moderator


Group: Moderators
Posts: 146
Member No.: 6
Joined: 7-July 02



QUOTE (vondrack @ Sep 10 2002, 10:05 PM)
BTW, SillKotscha... from what I have read on various sites, you should use precise bilinear for your resize. Use bicubic when scaling up, but bilinear when scaling down, they say... the picture should be better and the filter will run faster. Have you tried it?

thanks vondrack... I haven't tried it so far but I'll let you know about my results.

cheers Sill

--------------------
"Have you ever noticed that whenever Microsoft calls something 'Smart', it's definitely a feature you want to disable!"
 
     Top
blakerwry
Posted: Sep 13 2002, 02:02 AM


Unregistered









yeah, I agree with bilinear vs bicubic... bilinear when scaling down looks a little less blocky when compared to bicubic...

But by all means, use bicubic when upscaling.
 
  Top
bobsobol
Posted: Sep 16 2002, 11:20 PM


Unregistered









I have tried, as I know this is the recomendation. However (IMS) I found that you get flickering on thin, very light bands on a dark background or vise-vesa when the shot pan's in the vertical.

Say a shot of vanetian blinds maybe only a scanline or two high.

It's like interlacing.

The quality of Bi-Cubic is more blured but dose at least stop the flicker.

Try capturing the star-pan at the start of any Star Wars film at high res and try downscaling it with different levels of resampling from neerest neighbour to Presice-BiCubic. You will soon see what I mean no doubt.
 
  Top
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:
5 replies since Aug 30 2002, 04:10 PM Track this topic | Email this topic | Print this topic

<< Back to VirtualDub Filters and Filter Development