I noticed those before but never found the answer - why do they appear ?
Is it because of the color sampling ?
I don't think it is a result of de-interlacing since it looks like the artefacts exist even before the de-interlacing.
I will post here 3 links to losslessly saved screenshots:
1. The original before deinterlacing
2. After deinterlacing with QTGMC (parts with artefacts are marked with circles/ellipses, but there are many more unmarked as well)
3. Same as #2 just without the marks
Any ideas what is the reason for that ?
Thanks a lot !
+ Reply to Thread
Results 1 to 9 of 9
Last edited by Umen Pich; 23rd Sep 2012 at 15:08.Best wishes,
Yes, the same problem is in the "before" picture. It's just harder to see because of the comb artifacts. You should post a bit of the source so we can see if it's in the video or a byproduct of the way you're viewing the video.
here is a link to a few seconds of the raw material.
IMO the defect is most visible in the blue balls.
In general I've seen it in many other scenes but still cannot catch the pattern when does it appear most prominently.
Most logical seems - in cases of sharp colored edges, but I repeat this is just a first guess.
Thanks !Best wishes,
Your sample is uncompressed RGB. Is that the original or has it been converted from something else? A DV camcorder? YUY2 video capture?
It has a few problems. First there are dot crawl artifacts, though not consistently -- maybe composite video captured with a decent comb filter. On top of that it looks like it has some chroma sub sampling problems. Something like 4:1:1 helper may clean that up. But first I need to see something closer to the source -- assuming this has been converted from something else to RGB.
A quick trick you can do in AviSynth:
If you're not familiar with AviSynth you can get something similar in VirtualDub (or some other editor). Resize to 360x576 with a precise bilinear filter, then resize back to 720x576 with a precise bicubic filter.
But if we can get closer to the real source we can do better.
Attached is a short clip (h.264 mkv) of the above script plus QTGMC().
Last edited by jagabo; 20th Dec 2011 at 23:18.
Yes, the cause of the vertical stripes you are seeing are the dot crawl artifacts in the DV AVI. The fact that we don't see more of them (and sometimes there are vertical stripes instead of dots) indicates to me that you are using a capture device with a comb filter that's removing most of them. If you can disable that comb filter you might try capturing without it and using software filters like CheckMate in AviSynth, or Dot Crawl Comb Filter in VirtualDub. But I don't think those will work any better. They'll probably leave the same artifacts unless you crank the settings really high where they'll start screwing up the rest of the picture.
About the only thing I could come up with was the blur/sharpen trick. You can try playing around with the amount of sharpening but if you go higher you start to get oversharpening halos.
I don't use a capture device since the original comes from a digital camera. I use the little program WinDV to transfer from the camera to the PC. The second clip (from my prev. msg) is a cut from the DV file transferred that way. Absolutely no processing, no filters, nothing.Best wishes,
I thought maybe you were capturing a composite source with a DV device. But when I take your DV AVI, run it thought the blur/sharpen script (eliminating most of the dot artifacts), then reencode it as DV AVI again with Cedocida, some of the dot artifacts come back. So some of it is DCT ringing, not dot crawl. Some may also be the DV encoder in the camera or limitations of the photo sensors.