Horizontal lines/scanlines in captures? (ATI 600 USB)

[Turok81]

Why am I getting these horizontal lines in my captures? They are very subtle, but I notice them in all my captured tapes. Is this some kind of interlaced effect? They kind of look like scanlines to me. I also see a lot of combing artifacts(I think that's what it's called?) around moving objects, but I understand that this is normal so I'm not worried about that. I know you shouldn't deinterlace videos before putting them on dvds, but I deinterlacing a video with virtualdub just to see what effect it would have. It removed the combing as expected, but it had no effect on these horizontal lines in the background. They aren't that visible most of the time, but you can see them clearly in footage with blue sky. I've attached a couple of sample clips. I'm just wondering if it is normal, or if it is some effect produced by my ATI 600 device.

[sanlyn]

The disturbance isn't scanlines, it's Hanover bars and is common with PAL VHS. Sorry to say it, but your samples aren't usable as a cleanup demo, as they've been converted to uncompressed RGB. Hopefully you aren't capturing them that way. Poor control of input levels during capture is obvious with crushed shadows and clipped brights. All these problems are more effectively addressed in YUV before any other conversions, and with capture using proper level control. None of this is caused by your capture device. One of the samples has seriously bad halo effects and appears to be a multigenerational tape copy with numerous problems, including oversharpening effects.

If your samples had been captured and saved as losslessly compressed YUY2 their size would have been reduced to 1/3 the posted size, or about 34MB each.

[Turok81]

Ok, so they are called Hanover bars. So this is something that's part of the original tapes? So far I've only been capturing without using any compression codecs. I'm still at an early stage and I have only been capturing for the sake of testing and sorting out any issues. I figured I would use Huffyuv YUY2 when I'm ready to start serious capture/archiving. So you are saying the files are too large for a cleanup demo?

And yeah, one of the samples is from a multigenerational tape (the original is 8mm film)so it has some pretty bad bleeding and various issues.

-- merged --

I re-uploaded the samples. This time I saved them in YUY2 color space(that's how they were captured) and compressed them with Huffyuv. I wasn't aware the previous samples had been converted to RGB by default.

By the way, I had some difficulties getting Huffyuv to work. I installed both standard, 64bit and multithreaded version before virtualdub would recognize it. Yes, I rebooted my pc each time. The 64 bit version made my pc freeze and seemed to mess up a video driver or something. Using windows 10. Got a lot of flickering or no picture at all when watching some youtube videos. It seems to have fixed itself now though. Strange. Might try restoring os to an earlier date and install multithreaded Huffyuv again.

[sanlyn]

Quote:

Originally Posted by Turok81

Ok, so they are called Hanover bars. So this is something that's part of the original tapes? So far I've only been capturing without using any compression codecs. I'm still at an early stage and I have only been capturing for the sake of testing and sorting out any issues. I figured I would use Huffyuv YUY2 when I'm ready to start serious capture/archiving. So you are saying the files are too large for a cleanup demo?

No, the samples were not entirely suitable because of the RGB conversion, not because of their size. RGB is often preferred for specific work such as color correction, but a great many video defects are best repaired in YUV, which is the way most standard video formats are stored. You aren't using lossless compression during capture yet, but at least you captured using YUY2 rather than RGB. That's especially applicable to controlling video levels during capture.

Quote:

Originally Posted by Turok81

I re-uploaded the samples. This time I saved them in YUY2 color space(that's how they were captured) and compressed them with Huffyuv. I wasn't aware the previous samples had been converted to RGB by default.

Thank you for the new samples and the capture info. VirtualDub outputs RGB24 by default. If you want to preserve the original colorspace and compression when cutting a sample, use VirtualDub's "direct stream copy" mode instead of "full processing mode". If you want to keep the same colorspace but change compression, set the colorspace and compression you want and then use "fast recompress" mode.

You should use 32-bit VirtualDub and 32-bit add-ons, as well as 32-bit Avisynth when required. Filters and plugins don't necessarily run faster in 64-bit, they are simply able to access larger memory addresses. Also, you'll find that there are literally hundreds of 32-bit plugins and add-ons available but very little in 64-bit form.

Huffyuv-MT is obsolete and has nothing to do with multi-core processing, which is not the same thing as multithreading. The newest 32-bit huffyuv from their download site can automatically adjust to MT if needed. You'll also find that a lossless codec such as Lagarith or UT Video Codec will be needed for some restoration work because they can compress video in colorspaces other than YUY2 and RGB.

Examples of what would be possible with postprocessing of your samples are attached as .mpg's.

Your sample b kassett 2 test 4.00.avi looks a little dim, so I raised gamma by about 3 points in the attached post-process sample, to add some clarity to the figure against the trees. There are bad bright edge halos from over sharpening, and the main cable line is aliased. I used Avisynth to correct for these problems at a mild level that wouldn't destroy detail.

The sample b test 8.00b.00.avi has obvious problems, not all of which can be repaired. I worked on color balance and contrast, removed some spots, and moderated the edge ghosting. But those edge halos would require further experimentation. Likely it can't be 100% fixed without creating more damage elsewhere. The "wiggles" in verticals were caused by duping to tape without a line-level tbc in circuit and can't be removed, but I used QTGMC to smooth the wiggles a little. I don't know if the gamma boost near the end of the sample was in the original tape or not. The two frame hops around frame 66 could be repaired, but I didn't address them this time around.

The Avisynth scrips I used are attached as .txt files. The b_test8 sample required two steps, so I included a Step1 and Step2 text file. Whether you decide to use Avisynth or not is up to you, but I'd suggest that many of these image changes aren't available with VirtualDub or typical NLE edit controls.

[Turok81]

Ok. Thanks a lot for the cleanup demos. That's quiet a difference. It's good to know that all these options are available and that they really make a difference. I'm planning to learn how to use avisynth when I'm done capturing. Sounds like I need to tweak the settings in video levels. I haven't touched them yet, just left them as default since they looked ok to my untrained eyes. I've attached a picture of the settings. If I'm getting over-sharpening that probably means sharpness should be set to zero, right? I will check the histogram if I need to adjust the brightness and contrast.

[sanlyn]

Actually your capture levels were exactly within y=16-235 range. It was the generational copy that made it look strange, but that's because darks were crushed during the dupe, not during capture. A YUV histogram proved I was misled on that one, but it sure looked out of spec, LOL! Just goes to show you, trust the histogram, not eyesight.

In the histogram of the frame below, the white graph along the top of the histogram shows everything in the safe 16-235 range. The bunched-up spike of white at the left-hand side of the histogram shows darks crushed before capture, not during capture. The histogram is an Avisynth "Levels" built-in function.

[sanlyn]

Quote:

Originally Posted by Turok81

Sounds like I need to tweak the settings in video levels. I haven't touched them yet, just left them as default since they looked ok to my untrained eyes. I've attached a picture of the settings. If I'm getting over-sharpening that probably means sharpness should be set to zero, right? I will check the histogram if I need to adjust the brightness and contrast.

Sorry, I looked at your picture and was interrupted responding to it.

The kind of over sharpening halos I see appear to come from the VCR. If there is a sharpness adjustment on the player, it should be used first, usually at default position rather than turned all the way down. The capture proc amp setting usually stays at default, neither sharpening nor softening the image. If there's no sharpness adjustment on the player, you're a little stuck with the proc amp sharpener because it won't reduce sharpening artifacts already coming from the player, it just softens halos and everything else. In that case you have to do the best you can with post process filters rather than work with a blurry mess.

[Turok81]

Yeah, the sharpening is probably from the VCR. I hear that Panasonic decks tend to do that. Most of my tapes are VHS-C and that's the main reason I went with a Panasonic instead of a JVC. My unit is a Panasonic NV-HS950 and it doesn't appear to have any sharpness controls at all. So I guess it's just something I have to live with and try to improve in the post processing stage. Probably best to leave the video levels at default then. Thanks again for your input

[sanlyn]

Quote:

Originally Posted by Turok81

Panasonic NV-HS950 and it doesn't appear to have any sharpness controls at all. So I guess it's just something I have to live with and try to improve in the post processing stage. Probably best to leave the video levels at default then. Thanks again for your input

I have a similar Panasonic NTSC that I often use because of tracking issues, so I do what I can later. Incoming video levels are a different matter. Tapes will differ, and the only way to accurately prevent clipping from illegal levels is to use a capture histogram like the one in VirtualDub capture. Its use is pictured and described in Section 3 of a recent VirtualDub capture guide, in this post: http://www.digitalfaq.com/forum/vide...html#post45238. Scroll down in that post a few paragraphs to the screen captures of bad/favorable histogram adjustments via the "Levels..." filters.