Some guidance for improving VHS captures?
 

I tried many different pieces of software to capture some some old VHS tapes. I never could consistently get Virtualdub to capture without audio sync issues. Amarectv does the job perfectly. Now that I know how to capture without any audio sync issues, I'm figuring out what to do next.

I'm using the Roxio VHS to DVD Plus 3 USB stick and a basic 4 head VCR with a composite connection. I'm capturing at 29.97 fps, 720x480i with huffyuv video and PCM audio. I will not be spending money on a TBC, other VCR, etc. I know that is not ideal, but it is just the way it is and I'm OK with that. I want to make the video as good as possible without spending anymore money. I know the basics of using Virtualdub, Avisynth, Megui, Handbrake, and some other common tools.

I'm thinking I first need to calibrate my PC monitor. It is a HP 2009M. What is a good site or guide to use? I know doing it by eye is not ideal, but that is OK.

Next, I'm thinking I need to make sure that the colors are being captured as accurately as possible and/or use filters to correct after the capture. I need to learn how to make sure of that and how to make adjustments. I found this site http://www.trevlac.us/colorCorrection/colorTools.html Is it a good guide for doing that? I have briefly scanned it and will more carefully go over when I get a chance.

I also want to mask the edges of the video that would normally be covered by overscan. I know how to make an avs script with crop and addborders to do this. I can either do this while encoding with MeGui or I can run it thru VDub (adding any other filters needed) and resave it as a lossless huffyuv and then encode with x264 or maybe as a high bitrate MPEG-2. I know I could just crop, but I'm trying to maintain the full 720x480 frame and use anamorphic to display at 4:3 like needed. Am I going about it right?

I also am thinking about using QTGMC to deinterlace and denoise. I need to learn how to use it first. Any guides that are recommended?

Is there anything I'm missing or anything I'm not thinking right about? Thank you for the help.

Edit:

I have attached a short sample. I have been an amateur magician for around 25 years. When younger, I would record everything to do with magic that came on TV. I have over 100 hours VHS to capture. This will all just be for my personal collection. I know it will take a while, but that is perfectly ok. I also know some things may not be able to be fixed, but I want to fix all I can.

Welcome to digitalfaq. Thanks for the sample.

VirtualDub is sensitive to frame timing and line sync, which almost always results in audio sync issues. If you don't want a tbc, you have to workaround the obvious problems and live with the results. VCR's may not always require a frame level external tbc, but the noisy wiggling on verticals and angular lines, and the wiggles and notches in your side borders, can be avoided only with a line-level tbc at the time of capture. Much of the tiny jittery ripples in the video can be avoided with a line level tbc. None of these distortions can be repaired later. The playback shown is as smooth as you'll get without the recommended components.

Without better equipment, your sample in most respects is as good as it will get. The Roxio device is not what we would recommend, but without a better VCR a better capture device would be wasted. It's a shame you had to take a quality hit by using composite input, but I guess that's the way it is if you can't get better.

Do those ghost and double-images reflect the way the original program was broadcast? There's nothing one can do about those problems. If they're the fault of your VCR (which is either not tracking properly or is trying to track damaged tape) then you can't get a better capture. The white horizontal comets across the top look like permanent damage with bad tape heads. Notice that those comets also cover the left black border.

Yes. Do the crop and border restore with Avisynth. Don't even try to resize this video to fill the frame, as many incorrectly do. You can use crop and AddBorders to also eliminate the head-switching noise at the bottom. I don't think you can do it properly in MEGUI unless you alter its Avisynth script.

Colors look OK to me. But highlights are clipped (look at the top of the talking figure's white hair, which exceeds RGB 255 and smashes against the right-hand side of histograms). There's not much one could do about the blown-out white shirt in the fist part of the video -- much bright detail was likely burned out a long time ago, but some of the burn-out occurred during capture.
Levels are best controlled during capture by using brightness (for blacks) and contrast (for brights). This takes some fiddling, as both controls interact a bit, but you'll get the hang of it. If your capture software won't let you adjust levels, you're stuck. You can do what you can to fix crushed darks (I didn't see very much of that) and blown-out brights later, but it has to be done in YUV before doing anything else. Trying to correct color during capture with analog tape is an exercise in futility. Color Balance and even levels will change with almost every scene.

Note that ColorTools is a histogram/vectorscope. It doesn't adjust anything. It just tells you what's going on. It works only in RGB, so you can't use it during capture. Still, histograms are essential: no one's eyes are so good that they can tell what's happening with levels and colors, especially on a PC. Invaluable Virtualdub color filters would be ColorMill (http://fdump.narod.ru/rgb.htm) and gradation curves (http://members.chello.at/nagiller/vdub/index.html). The latter has a tutorial (http://members.chello.at/nagiller/vd.../tutorial.html). Gradation curves are similar to the curves filters in Vegas Pro, Premiere Pro, After Effects, FCP, and other pricey apps. You wont find filters like these in budget NLE's or the encoding apps you mentioned.

I'd suggest that you dig deeper into the basics. Your video sample isn't interlaced. It's telecined. Deinterlacers don't work with telecined video. It was originally produced on film at 23.976fps, then telecined for broadcast at 29.97fps. Neither MEGUI nor Handbrake perform inverse telecine properly -- in fact, Premiere Pro doesn't do it right, either, it just makes a mess. Avisynth has several inverse telecine filters, the most popular being TIVTC. When telecine effects are removed, this video runs at 23.976 progressive. If you want DVD or standard definition BluRay, restore telecine during encoding -- in this case with 3:2 pulldown. For denoising, almost any filter you use requires telecine to be removed first.

A problem in your sample is the first segment with the boy. It appears to have been created as 18fps or 20fps movie film, then with duplicate frames during optical transfer to some other format (using a combo of duplicate and blended frames). If you just remove most of the duped frames, the video runs at 25fps and won't sync or join with the 23.976 segments. Your best bet is to IVTC that segment with the rest of the video and accept some duplicate or blended frames. The blended frames look field-blended, which can't be corrected anyway after all the pre-processing that the original video went through.

With true interlaced video, QTGMC is the only choice if you want anything that doesn't look damaged (limited, of course, by the condition of the source). With progressive video QTGMC can be used as a denoiser with InputType=1 or 2. There's no dedicated tutorial for QTGMC. People pick up tips for it in restoration forums, although it does have a dedicated doom9 discussion thread (http://forum.doom9.org/showthread.php?t=156028)) which has over 100 web pages. Its documentation is the html that comes with the plugin package -- helpful, even if at least 50% is incomprehensible but it's a start for most of us. Usually you can use QTGMC's presets, which are explained in more detail in the top couple of hundred text lines of the avsi itself. That avsi is text-formatted: open it in Wordpad, not in Notepad, and save a ".txt" copy for yourself for more convenient access. Don't use WordWrap in Notepad when reading or saving the file. And don't make text changes in the original avsi or you'll have serious trouble. This complexity in settings often isn't necessary; most of the time, just use presets ("slow" and "medium" are favorites. Slow is the default but can often over filter). MCTemporalDenoise (MCTD) is a similarly convoluted filter -- people just use its own presets as well.

For true interlaced video you can do tests a lot quicker using yadif (http://avisynth.nl/index.php/Yadif) in mode 1. Yadif isn't nearly as clean or sharp as QTGMC (QTGMC does have a repair mode that loads yadif for deinterlacing), but with slow QTGMC speeds a script in VirtualDub or other editor can be slow and/or stuttery. Many people use a yadif quickie just to see what's going on inside a video, then replace it with QTGMC for the final script. You don't always have to deinterlace to denoise, but sometimes you have no choice. For most standard output formats, deinterlaced video should be reinterlaced in Avisynth, or telecine restored with your encoder. For PC playback and a few external players, many can work with progressive video. But standard definition DVD or BluRay/AVCHD isn't encoded as progressive. Some of BluRay's various formats (for instance, 1920x1080 or 720x480, @29.97fps) are not progressive.

Well, LOL, I wouldn't just think about it. Just do it. I don't see how people can work with video on uncalibrated monitors, but they do (and the results show it). Manual calibration of these critters is nigh near impossible if not difficult, but you can improve some areas. HP consumer monitors don't generally have a gamma setting, so you'll have to do it with brightness and contrast controls and some test patches. You can find some at this popular site: http://www.lagom.nl/lcd-test/. Really accurate color adjustment isn't possible with monitor image controls, which are incredibly basic and aren't linear. But the lagom site is better than nothing.

True monitor calibration is done with colorimeters and associated software that create hardware profiles and LUT's (Look Up Table matrices) for graphics cards. XRite and Spyder are the usual tools, with XRite preferred. But they aren't free. And don't buy them used -- colorimeters become inaccurate with time. Obviously you don't want to spend on this, but if you're curious here's a test report and review to show exactly how it's done, in mind-boggling detail (with pictures to help): http://www.tftcentral.co.uk/reviews/...e_display2.htm. It's a revealing review that shows you how un-calibrated an expensive monitor like a Dell UltraSharp can be out of the box.

Most people get greater color accuracy with IPS monitors instead of TFT LCD's. But that's yet another expense.

Using histograms to check levels, saturation, color balance, etc.:

The image below is a frame from your sample, shown after inverse telecine was removed. Under the image are examples of histograms and waveforms taken on this frame, in YUV and then in RGB. In your telecined sample this is frame 217. After telecine is removed, it becomes frame 174.
http://www.digitalfaq.com/forum/atta...1&d=1447092625

Below: a YUV levels histogram (left) and a YUV waveform (right). In the left-hand histogram notice the top white area, showing how luma levels exceed RGB 16-235 in the shaded border at the right side of the graph. In RGB, the discolored shaded area will be clipped. The right-hand waveform is vertical instead of horizontal, but it shows the same effect along the right-hand side. Black borders were removed from the frame to avoid affecting the histogram. Black borders will be a "spike" in the darker left edge of both graphs.
http://www.digitalfaq.com/forum/atta...1&d=1447092701

Below: RGB histogram and waveform from the same frame. In RGB, dark and bright limits of YUV colorspaces are expanded at both ends. The darks (left-hand side) aren't in bad shape, but the brights and color saturation crash into the right-hand side, indicating clipping -- too much contrast in luma and chroma alike. The RGB graphs were made in VirtualDub with ColorTools. In the waveform at the right, bright clipping is shown smashed against the top of the graphs.
http://www.digitalfaq.com/forum/atta...1&d=1447092760

Levels and chroma should first be corrected in YUV. Once they get clipped in RGB or in TV display, it's too late. The bright clipping could have been avoided by adjusting contrast during capture. Even here, however, if YUV levels are corrected in Avisynth you can actually see some detail in that bright shirt.

With the video as-is, there's not much one can do with mpst of the defects. A few things can be repaired. Below are three consecutive frames with telecine removed. In the middle frame you see a big dropout or "rip" right across the midsection of the image. In your original sample, this is frame 217 again.
http://www.digitalfaq.com/forum/atta...1&d=1447092872

Below, the middle frame is a new interpolated frame using ReplaceFramesMC in Avisynth. The correction should be made with progressive (deinterlaced or inverse telecined) frames.
http://www.digitalfaq.com/forum/atta...1&d=1447092918

Considering the setup used and the tape itself, you got more than most beginners would. But unless your tape is seriuously damaged, this is not a good capture. Without better equipment as usually recommended here and elsewhere, you won't get cleaner or better captures than what you have. You might consider taking one or two of your most valued tapes and sending them to a pro capture shop such as digitalfaq's services for a good lossless capture. I don't work for those guys but I used their services on an occasion when my prime VCR went on the blitz for a while. They know what they're doing. Considering what you'd have to spend and learn just to get better capture gear, it would be cheaper to have a pro shop capture those 100 hours for you. Everything you'd need to post-process is free.

After playign with filters for a while with this sample, I'd say it's no good treating one segment one way and the other segment another Way. You'd get unmatching frameb rates by do9i8ng this, or poor motion in at least one of the segments. You'll get cleaner results avoiding QTGMC or deinterlacers, instead denosing and denoising with filters that can work with SeparateFields (such as MVDegrain2 or dfttest) or have provision to set filtering for interlaced video (such as MCTD mentioned earlier).

Either way, don't expect very much. Most of the sample's problems that can't be repaired. What you really need is a better capture.

The most obvious issue I see is with the hardware. You're using a low-quality capture stick that's known to reduce color quality and resolution. It also tends to overexpose and blow out highlights. It's essentially the same POS as an Easycap/EZcap.

Couple that with a low-end consumer VHS VCR, and you've always lost.

Most of the video quality was lost before software could touch it, so Avisynth/etc is almost pointless now.