VideoHelp Forum
+ Reply to Thread
Results 1 to 15 of 15
Thread
  1. Member buckethead's Avatar
    Join Date: Jul 2008
    Location: United States
    Search Comp PM
    8mm and Super 8 films can have lots of color issues when restoring and transferring to the digital domain. I'm at the point in my learning curve where I have discovered that the MSU Old Color Restoration Filter (v 0.9) when used in virtualdubmod can find fairly accurate colors which can then be worked with.

    My only problem with this filter is that areas in the original with any luminance become VERY white with a loss of details. Here's an example with credit to themaster1 :

    http://img185.imageshack.us/img185/5454/image2vv3.jpg

    Although I can use Gradation Curves to lessen the luminance in the top range, the details are still lost.

    Can anyone offer any help or suggestions ?
    Quote Quote  
  2. Member
    Join Date: Jul 2001
    Location: NY
    Search Comp PM
    I am sure you will get a professional answer on here, but perhaps toying with the gamma (some filter must exist for this) might help you out a bit

    PS: thank you for the heads up on the old color filter
    Quote Quote  
  3. Member themaster1's Avatar
    Join Date: Nov 2006
    Location: France
    Search Comp PM
    This bring us back to this thread

    color mill may help you to work on the levels (dark/middle/light)
    Quote Quote  
  4. Use AviSynth to open your source then add ConvertToRGB(matrix="PC.601"). That will give you more head (and tail) room. You might have to follow with Levels afterwards to fix the levels (blacks at RGB=0, whites at RGB=255).
    Quote Quote  
  5. Video Restorer lordsmurf's Avatar
    Join Date: Jun 2003
    Location: dFAQ.us/lordsmurf
    Search Comp PM
    Word of warning -- ColorMill has created interlacing damage on some videos I've used it on. It's very random, I don't know what causes it just yet!
    Quote Quote  
  6. Originally Posted by lordsmurf
    Word of warning -- ColorMill has created interlacing damage on some videos I've used it on. It's very random, I don't know what causes it just yet!
    Interlaced YV12 sources? VirtualDub doesn't handle interlaced YV12 properly. It treats it as progressive, thereby blending the chroma channels of the two fields. The workaround for this is to open your YV12 source (MPEG 1/2/4, PAL DV) with AviSynth and convert to YUY2 or RGB with the interlaced flag set to true:

    Code:
    WhateverSource("some YV12 source")
    ConvertToYUY2(interlaced=true)
    Of course, you'll need to follow that with interlace aware filters or techniques in VirtualDub.
    Quote Quote  
  7. Video Restorer lordsmurf's Avatar
    Join Date: Jun 2003
    Location: dFAQ.us/lordsmurf
    Search Comp PM
    Nope, not YV12.
    It was an I-frame only high bitrate MPEG-2 opened with MPEG2.vdplugin in VirtualDub 1.8

    I'm not convinced ColorMill is an interlace-aware filter!
    Quote Quote  
  8. Originally Posted by lordsmurf
    Nope, not YV12.
    It was an I-frame only high bitrate MPEG-2 opened with MPEG2.vdplugin in VirtualDub 1.8
    That would be YV12. The MPEG decoder decodes as interlaced YV12, then VirtualDub converts to RGB incorrectly. Here's are some samples with VirtualDubMod, VirtualDubMPEG2, and VirtualDub. VirtualDub handles the interlaced MPG data incorrectly:

    Quote Quote  
  9. Video Restorer lordsmurf's Avatar
    Join Date: Jun 2003
    Location: dFAQ.us/lordsmurf
    Search Comp PM
    Well, sumbitch...
    I'll need to re-add the MPEG2 and MOD to the system, if that's the case.

    What I observed I cannot explain. It was not field order issues, but there was some sort of oddity going on, move clearly viewed during a pan.

    What has thrown me, however, is the fact that it was random. Two other projects looked fine, only the one was all fubar.
    Quote Quote  
  10. I forgot to mention those screen caps were at 4x enlargement. The red lines are one pixel thick in the originals.

    And, as mentioned, you can use AviSynth along with Virtualdub to get around the problem.
    Quote Quote  
  11. Video Restorer lordsmurf's Avatar
    Join Date: Jun 2003
    Location: dFAQ.us/lordsmurf
    Search Comp PM
    Is there a way this can be repaired? I have a video here that had its interlacing messed up by the aforementioned VirtualDub (MPEG-2 source) YV12>RGB conversion, something cleaned up by Colormill.
    Quote Quote  
  12. What has been done to the video? You opened the MPEG 2 video in VirtualDub -- and then what? Some filtering or conversions might leave the video still fixable. Can you provide a short sample?
    Quote Quote  
  13. Video Restorer lordsmurf's Avatar
    Join Date: Jun 2003
    Location: dFAQ.us/lordsmurf
    Search Comp PM
    Well, I ran the first one through Restream, and it was actually fine after reversing interlace. I'm going to see if I run across anything that Restream won't handle correctly. It's not as bad as your 1.9.2 image above.

    The footage is so degraded that it didn't look like a simple reversal, but that may be all that happened.

    I'll keep you informed...

    Thanks.
    Quote Quote  
  14. Originally Posted by lordsmurf
    It's not as bad as your 1.9.2 image above.
    That video was chosen as a very bad case example. Most real world video with the problem won't look as bad.

    Originally Posted by lordsmurf
    The footage is so degraded that it didn't look like a simple reversal, but that may be all that happened.
    It's not a field reversal. It's a case of interlaced chroma channels being handled as progressive chroma channels. In progressive YV12 the first line of chroma is applied to scanlines 0 and 1 of the luma channel, the second line of chroma is applied to scanlines 2 and 3, etc. In interlaced YV12 the first line of chroma is applied to scanlines 0 and 2, the second line of chroma is applied to lines 1 and 3, etc. That is, the chroma is interlaced as well as the luma.
    Quote Quote  
  15. Video Restorer lordsmurf's Avatar
    Join Date: Jun 2003
    Location: dFAQ.us/lordsmurf
    Search Comp PM
    Well, mine is appearing to be a simple reversal --- that's what's so confusing about this.
    Quote Quote  



Similar Threads