Welcome to Doom9's Forum, THE in-place to be for everyone interested in DVD conversion. Before you start posting please read the forum rules. By posting to this forum you agree to abide by the rules. |
|
|
Thread Tools | Search this Thread | Display Modes |
20th October 2011, 15:41 | #43 | Link | |
Avisynth language lover
Join Date: Dec 2007
Location: Spain
Posts: 3,431
|
Quote:
|
|
20th October 2011, 16:28 | #46 | Link |
Registered User
Join Date: Jan 2006
Posts: 1,867
|
I used the version I posted and it works, do you have GRunT installed as well? I think I'll just remove the dependency on Gscript/Grunt...
New comparison http://screenshotcomparison.com/comparison/88873 This shows lining up with video only. To me it looks like it's worse. So that means my hsync technique does help and the side should look ragged when TBC'd. |
20th October 2011, 17:16 | #47 | Link | |
Registered User
Join Date: Oct 2011
Location: Germany
Posts: 39
|
@Gavino
Thanks! Works with Avisynth 2.5.8 and GRunT installed. Quote:
If the black inside the line cannot be separated from the black of the border perhaps some plausibility checks must be involved (see this post). Last edited by sven_x; 20th October 2011 at 17:30. |
|
20th October 2011, 17:57 | #48 | Link |
Registered User
Join Date: Jan 2006
Posts: 1,867
|
Aha. There's other differences, but I can't tell which is better. I should run a test signal through. I left detection as a mask just for the purpose of problems like this. The resulting shift mask can just be blurred to reduce sudden changes! The same ideas of limiting it can be turned into a vertical convolution...
How does it look on your sources? |
20th October 2011, 18:43 | #49 | Link |
Registered User
Join Date: Feb 2002
Location: California
Posts: 2,691
|
I just tried the script, unaltered, on some 720x480 NTSC interlaced footage. The left side is the original footage and the right side is the footage processed by the script. As you can see, it looks like the script is actually producing the very problems it is supposed to eliminate! I didn't alter anything in the script -- just downloaded it and ran it.
Do I need to have the additional information that your capture drivers provide? If so, how do I get that? Last edited by johnmeyer; 20th October 2011 at 18:44. Reason: Forgot to add the last line. |
20th October 2011, 19:42 | #50 | Link |
Registered User
Join Date: Oct 2011
Location: Germany
Posts: 39
|
@johnmeyer
Did you feed the script with the left input? So there are the borders missing that the script uses to find the transition to the beginning of the "real" line beginning and end. Without borders it cannot estimate anything. @jmac698 My sources are grabbed with an 15 Euro USB Video Grabber and a fairly good, very old Toshiba VHS recorder. With this setup sync disturbances are rather small. In most cases your script does not find another solution to line up the lines. I can see no differences between source and processed (switching with AvsP through the scripts). One video is the copy of a copy from a camcorder. This is the only one with some ripple at the border that needs adjustment. The result is better. In one case the result looks pure: lots of lines out of row that were okay in the source. I have no idea what is causing this. The borders look clear and smooth in the source. PAL recordings have 576 lines. A part of the algorithm seem to work with 480 lines only. I have the impression that removing the blur effect (that occurs when merging two frames) requires sub-pixel accuracy (at least 1/2 pixel). Last edited by sven_x; 21st October 2011 at 17:13. |
21st October 2011, 01:14 | #51 | Link | |
Registered User
Join Date: Feb 2002
Location: California
Posts: 2,691
|
Quote:
I've read through the script, and tried changing both the "thresh" and "searchwidth" parameters, but that didn't fix the problem. The left border in the original (before) is about ten pixels wide, and the black border on the right is about four pixels. Here's a link to 2-3 seconds of the footage from which the left snap was taken: Test File |
|
21st October 2011, 02:09 | #52 | Link |
Registered User
Join Date: Jan 2006
Posts: 1,867
|
Top post updated. It's up to you to make a mask, try returning mt_binarize(thresh) to see what's going on. AvsPmod is useful to cursor over the black bits and read the Y (luma) value. It was made for a special capture with hsync and/or viewing blacker than black at the borders (with NTSC-J capture).
|
21st October 2011, 17:18 | #55 | Link |
Registered User
Join Date: Oct 2011
Location: Germany
Posts: 39
|
@johnmeyer
Opening your source screenshot in Photoshop, selecting the inner image area and displaying a histogramm shows, that the image possibly has undergone a levels adjustment. The levels of the inner area lay in the range 0 ... 255 (well, deepest luma is a bit higher), so that searching for a minium might not land in the border area. After a levels adjustment the borders are not in a deeper black. Nevertheless in one case the script produced also a bad result with my own recordings. That was recorded from TV with a very cheap PAL VHS recorder. Last edited by sven_x; 21st October 2011 at 17:41. |
21st October 2011, 18:29 | #56 | Link |
Registered User
Join Date: Oct 2010
Posts: 18
|
Iīm trying to use your script but Iīm getting this error message:
"Script error: there is no function named "mt_binarize" ([ScriptClip], line 5)" I have Masktools v.2a48 installed (also GRunT and GScript). Iīm using Avisinth 2.6. Iīve looked for a solution, but didnīt find it. Donīt know what to do. I know probably this is a noobie mistake, but I appreciate if somebody could help me... |
21st October 2011, 19:13 | #57 | Link |
Registered User
Join Date: Jan 2006
Posts: 1,867
|
That doesn't seem a problem with the script itself, but you can try
Code:
colorbars(pixel_type="YV12") mt_binarize(162) |
Tags |
tbc |
Thread Tools | Search this Thread |
Display Modes | |
|
|