|
|
| Quilty |
| Posted: Jun 18 2003, 11:12 AM |
 |
|
Unregistered

|
 Here you can see the problem .... does anyone have a filter to correct this occurance? It seems to be only on thin lines of very saturated colours. The "staircase" pattern alternates from frame to frame (as I have simulated in the GIF). This has been through a Kramer Comp to S-video convertor, so the rest of it looks much better than the composite original, save for this error. I have tried the 2d comb filter, and Guava comb filter, and whilst Guava cleans it to a great extent, it mucks up the rest of the image.
There is a reward for the person who fixes it!!
Cheers,
Quilty |
 |
| tereza |
| Posted: Jun 18 2003, 02:07 PM |
 |
|
Unregistered

|
have u tried some deinterlace filters? Tom |
 |
| Quilty |
| Posted: Jun 18 2003, 02:15 PM |
 |
|
Unregistered

|
Sorry .. should have pointed out these are progressive frames Inverse Telecinied back to 23.976fps in After Effects. Thanks for your response, but I am not sure that would work 
Q. |
 |
| phaeron |
| Posted: Jun 19 2003, 02:50 AM |
 |
|

Virtualdub Developer
  
Group: Administrator
Posts: 7773
Member No.: 61
Joined: 30-July 02

|
Those are chroma crosstalk artifacts. NTSC/PAL use quadrature encoding to encode the two color axes into a single modulated signal; as it turns out, the bandwidth allotment for one of the axes is asymmetrical on the reddish-orange side. It looks like your S-Video converter uses a symmetrical chroma separation filter, resulting in saturated reds bleeding into luma. This isn't uncommon, particularly with computer-generated output (Amiga 1000, anyone?).
One way to combat this problem would be an adaptive low-pass luma filter that's keyed to the reddish-orange chroma region. Unfortunately, I don't know offhand of a direct way to do this. Any Avisynth wizards have suggestions?
Also, why are you using a converter? Converting to S-Video isn't going to gain you anything over capturing composite directly unless the external converter you're using has a better chroma separator, and that doesn't really seem to be the case....
|
 |