digitalFAQ.com Forum

digitalFAQ.com Forum (https://www.digitalfaq.com/forum/)
-   Restore, Filter, Improve Quality (https://www.digitalfaq.com/forum/video-restore/)
-   -   Good quality restore/capture setup? (https://www.digitalfaq.com/forum/video-restore/433-good-quality-restore.html)

tobias 03-18-2009 04:00 AM

Good quality restore/capture setup?
 
Hi lord,
To continue our conversation from the other forum:

- My DR-1000 works with PAL
Ok, I emailed SignVideo to ask for shipping and got the following reply: "I am afraid we no longer manufacture PAL versions of our PA's and/or Image Enhancers. Sorry. Only the NTSC version is currently produced.", so perhaps you managed to get a PAL version before it got discontinued? Any other image enhancer I should look into?

- My Elite Video BVP-4 Plus works with PAL
This looks like a suitable alternative, too bad I can't seem to find one on ebay. Any suggestions to where I can find one?

- PAL equiv of Panasonic AG-1980P is good.
OK, will go with one of these to start with.

- Matrox RTX cards are great, if you have the funds.
Indeed they were a bit pricey, but if they are worth the money..

- I don't really know your needs, hard to suggest. It depends on what you'll do with the video once captured (edit, dump to DVD, etc)
Well no editing will be done on the material. Will only restore the material, digitalize and burn to dvd. Good quality DVD-burners are already available in-house.

- You can't do more than one capture per computer at any time. All hell will break loose when your try to multi-capture. It's a bad idea.
Ok, understood.

Is there anything obvious that I'm missing/forgetting here now?

Best regards,
Tobias Lolax (peep)

lordsmurf 03-18-2009 04:13 AM

Are you planning to restore as much as possible in hardware only, or will you restore in software too? Some errors cannot be repaired in hardware only, such as excessive grain, or severa color damage. How damaged are the videos that you have now, or how "ready" must you be at all times to handle excessively damaged videos?

tobias 03-18-2009 04:45 AM

Well the more hardware restoration, the better the result and faster process, right? If software is needed to correct some of the errors, then that's the way it will have to be.
Not even the client knows the condition of all the tapes. They are quite old since it's all archive material, but they have been stored well so hopefully they will be in good condition. But we do need to be ready to handle videos of worse quality aswell. We do have baking equipment, and if my memory serves me correct, it can also handle VHS tapes.

lordsmurf 03-19-2009 01:43 AM

Yes and no. Some errors can only be effectively corrected in hardware (video chroma noise, for example), while others can only be corrected in software (audio noise, for example). It really depends on the error. Proc amps, for example, can only alter color so much. Software can go further. It's not uncommon that I'll start a restoration project in hardware, and then continue and finish it in software. Sometimes it can take multiple passes in software, to fix the color issues.

For some folks, that's too much work, they quit at the hardware stage. That's fine, if all you're doing is conversion, not advanced restoration work.

To be honest, age has very little to do with quality. That's such a myth, especially in recent years. The quality of the signal that was recorded, and the equipment which did the recording, are far more important aspects. Storage and use are secondary issues.

A baking setup is good to have. That's something I won't mess with, due to cost and learning curve. Aside from major hurricanes (such as the ones in the USA in 1995), there was never much business opportunity in it. For future consideration, if I come across clients in Europe that need tapes baked, would you want the referral business?

I might have an extra BVP-4 Plus available, if you're interested. These originally ran about $1,500 or more. This one has had very little use. Let me know if you'd like to buy it, and I can decide on a price (feel free to make an offer, too).

The only think you're really missing is the workflow. What is your proposed workflow, from source tape to final disc (or whatever the final format is), including any hardware and software steps?

tobias 03-19-2009 06:21 AM

I am indeed very much interrested, so please reply or PM with your best offer.

To be honest, I haven't even looked into much of the digital part yet (software/filters). Been too much reading up on the analog world of the restoring business. But the workflow as of now would be Source -> Pana VN-FS200 -> S-Video -> BVP-4 Plus(perhaps) -> S-Video -> Matrox RT.X2 SD.

Havent looked into what software is supplied with the Matrox package, but perhaps I'll use VirtualDub/VirtualVCR to capture it in a lossless format (eg with the Huffyuv codec), process/filter the material if needed (AviSynth/VirtualDub?), encode to MPEG and burn to DVD. Software-wise, I'm open to suggestions and would really appreciate your opinion on it. Been in contact with VirtualDub and Adobe Premiere/Encore/Audition/Soundbooth in other private projects, so I know of their capabilities.

lordsmurf 03-19-2009 06:41 AM

The Matrox cards are professional equipment, with drivers made for use in Adobe Premiere. I don't know that capturing in VirtualDub would be possible, although uncompressed AVI export from Premiere (or DV export) is possible. Then load that into VirtualDub, to take advantage of its restoration abilities.

My experience with Matrox was the pro-end card used in a Mac, OS9, some years ago, with Final Cut Pro. It worked well, for what it did.

For years, less-expensive "non-professional" capture card options existed for Windows/Linux boxes, but those days are now gone. If you're lucky, you can still sometimes get an ATI All In Wonder Radeon AGP 9x00 series card -- assuming your computer accepts AGP! I've kept onto AGP boxes for that reason, but two of my machines are now PCI-e, making a Matrox card the only really good solution, should I need a capture card in those machines.

The Matrox RT2500 is the lowest/oldest card I'd go with. To be completely honest, the Matrox RT.X100 card would probably be great too, although I'm not sure what version of Premiere it works with. I will say that Premiere CS3 is the best version of Premiere, if your machine has the horsepower (Intel box with dual-core, 2GB RAM or more, lots of hard drive space). I do have Premiere Pro 1.5 on my older AMD64, and Premiere 6.5 on my oldest Intel because neither machine is hefty enough for CS3. I can't get CS4 to be stable on anything I use right now, a common gripe at the Adobe forums.

For the Elite Video proc amp -- which is a LOVELY unit, by the way, I use one as do many of my associates -- I'd be looking for about $800 plus shipping. That's half price or less from MSRP. The lowest cost I ever saw on these units was about $500 five years, in abused condition. This one was bought for a specific purpose, used a short time, and then stuffed back in the shipping box for years. I had a great TBC like that too, sold it several months ago when somebody was looking for a good hard-to-find TBC.

We can go over some restoration techniques in the Restoration sub-forum on this site (also under DVD Projects area), when it comes time to talk about software filtering, some scenarios, etc. VirtualDub, even though it's freeware, is part of my weekly operations. It restores better than Premiere ever could.

If you've not used Sony SoundForge or Goldwave for audio restoration, it's worth a try. It sometimes makes the Adobe audio software look weeny.

lordsmurf 04-08-2009 02:46 AM

My primary workflow is as such:

S-VHS VCR > TBC > Proc Amp > Capture Device

VCRs: Panasonic AG-1980P or JVC HR-S9800 or JVC SR-V10U or JVC HR-S7965EK (PAL)
TBCs: AVT-8710 or DataVideo TBC-1000
Proc Amps: Elite Video BVP4+
Capture Device: various DVD recorders and capture cards


Sometimes the TBC is required prior to the proc amp. Proc amps can "freak out" if the signal is too noisy. The BVP4+ can be especially touchy, because of how aggressively it is able to augment a signal's color, black/white levels, contrast and resolution.

Other optional devices are available, as needed, such as a Panasoinc ES10 pass-through filter, or a Studio 1 Productions detailer. The inclusion location in the workflow depends on several variables (not discussed at this time, that's a whole different conversation, if needed).


More on the BVP+ proc amp...

When it comes to the BVP4+, it is possible to alter the signal too much, as it is powerful. One good example is aggressive IRE or luminance changes, which can destabilize the signal. These severe changes are often unneeded and ugly anyway, a marginal turn of the knobs, just a couple of increments either direction, is enough to correct the video clarity.

To increase contrast, increase the luma, and decrease the IRE -- or vice versa. Let your eyes be your guides. Just be sure your preview monitor is relatively close to neutral and calibrated decently. Consider using a Avia disc, if available.

Typically you'd never go more than +5 tick-marks on the color saturation boost. I find the "flesh tone" to be minimally useful, it depends on the tape. In most cases, I'll just opt for the full hue change, either +180 or -180, to alter color quality from poorer tapes. I rarely use the black restore. In general, it's not a good idea to use the resolution boost more than a "half turn" (white marker aimed upwards). If you need more sharpening, then enable VCR sharpening (the Panasonic is better than the JVC here), and/or use a detailer (such as the Studio 1 / SignVideo devices).

Remember that you probably cannot fix an overexposed or underexposed video, just make it less awful in quality.

The herringbone filter has been useful maybe once per year, not too often. It's very minimal, not aggressive at all. The knob on back is supposed to increase the bit-depth of color, depending on analog or digital source, but I've never visually perceived much difference.

If you need more advice on the workflow, or he BVP4+ itself, ask. It's really an easy device to use, especially if you have a even a basic understanding of analog video.

tobias 04-09-2009 08:18 AM

Thank you so much for the last post. How about software-wise? We still haven't discussed software filters and settings. I'll capture using Premiere, export with uncompressed avi and if needed run it through some filters in VDub. Any filters that you suggest/tend to use? Also, any suggestions on the conversion and authoring to/of mpeg (final destination = DVD)?

Just to let you know, I managed to extend the timeschedule a bit, so I now also have the weekend to complete the first batch. There were some delay on some of the equipment, the procamp arrived on wednesday and the new harddrive arrived just now. Just have to go get some power cables that allows me to connect the Pana and the procamp to the ups and install the harddrive and I'm ready to go.

lordsmurf 04-11-2009 02:35 AM

Hardware restoration tends to be a little more "generic" in the methodology. In some ways, it's limited, simply turning on a TBC or calibration, selecting one of a few options for audio and video filtering, tweaks a few dials on a proc amp.

Software restoration is very error based. There are probably 100 different scenarios I can think of right off the bat. Someday, I hope, this forum will act as a pretty good catalog of those errors, as more folks use the forum, asking for help with their videos. When that day comes, it will be a simple issue of linking to some topic and samples. In the meantime, however, it's a case-by-case basis. When you run into a certain error, post about it here, attaching a still-image sample, and then (in some way) getting us a video clip to view.

In general, broadly speaking, you'll often want to suppress grain noise, something done with VirtualDub or Deemon Video Enhancer (both use VirtualDub filters).

Masking and/or cropping sometimes enters the picture too, depending on the application of the video. This is done in any number of editors or encoders, including Premiere, VirtualDub, Procoder, MainConcept and TMPGEnc.

Deinterlacing -- a GOOD AND PROPER method -- is often important if you're creating professional-quality web streams. This is often achieved using the "Deinterlace Area Based" filter for VirtualDub.

Further color corrections are performed in Premiere or, my favorite, TMPGEnc Plus.

Further chroma removal (beyond what the hardware could do) can be minimally processed in VirtualDub too. Chroma noise isn't as much of an issue with PAL video as it is NTSC video, so you may be lucky here. Chroma noise is an evil menace for us, sometimes.

Herringbone noise is often filtered the same as grain, but with much harsher temporal filtering.

Up-resolution of weak-quality formats can be performed in one of several ways, in VirtualDub or TMPGEnc Plus. I doubt you'll run into this, if your projects are analog sources, and not already-butchered digital sources.

De-shaking unsteady video can be worked on in VirtualDub and/or SteadyHand. This is by far one of the more challenging restoration tasks.

Audio errors are another area best dealt with on a case-by-case basis. There are some guides here in this forum, and on the site itself (HTML web page guides). SoundForge and Goldwave are all I really need, with the ability to remove hiss and buzz, correct tonal irregularities, and just generally make it sound better. I've worked with digital audio for at least 15 years now.

I'll gladly take a sample clip sent to me, run it through VirtualDub, let you know what filters I used (and attach needed filters to the post), and then upload my VCF settings file, so you can just load it up and process your full-length videos.

Audio restoration can be done much the same way, where I fix the audio, process/restore it, and then return a sample to you, along with instructions on how it was done. That way you can restore the full-length version, and then learn by doing.

The Adobe Media Encoder found in Adobe Premiere CS3 and CS4, which uses the MainConcept engine, is easily one of the best quality encoders right now. So use that for export. If you need help with the tweaking of settings (bitrate, etc), let me know.

tobias 04-11-2009 02:55 PM

I cut and exported a compilation taken from the first vhs-capture. Note that this is not compressed so the filesize if rather large (700mb/30sec). Sadly the RT.X2 cannot capture video noncompressed so the clip is captured using Matrox MPEG-2-I-frame settings (25Mb/sec data rate - the maximum). The material is not edited in any way, haven't cropped it yet or anything. I took some clips I thought could be filtered a bit (noise/bleed/..).

The procamp was setup to pass on the color adjustments, color level close to 0, PTP 0, IRE +2 and quite a lot on the resolution boost. I increased the resolution to that point that the ratio between noise generated and detail gained was "suitable"? But since I will now have to filter out the noise generated, all of the detail gained might go lost anyway? What would you say about the levels (color/black)?

I had to install Vista on the workstation, due to the previous OS not being supported by the Matrox software, and therefore my screen is no longer calibrated (at least not in vista, need to first capture a short clip and then boot my other OS to check the color settings). I do have the TV set up to the output of the breakout box but the colors aren't calibrated on that input with s-video either. I won't have access to the calibrator until after the weekend so I will just have to manage over the weekend.

The clip can be downloaded from http://www.kcproduction.fi/compilation.rar

Thoughts / suggestions? Time is getting tight even though the deadline was postponed, too much unexpected things has come up. To add to all this mess, I've gotten a really bad cold and is caughing my lungs out.

I hope you have time to have a go at this.

Br
Tobias

lordsmurf 04-13-2009 01:13 PM

3 Attachment(s)
I was able to download the file, and do some quick testing. What I observe on the video is chaotic noise in the analog signal, not an uncommon issue. This is quite easy to fix.

You are correct that applying certain kinds of filters would negate the sharpening (resolution boosting) that you did via the proc amp. An intra-frame type noise filter would indeed tend to soften the image.

What I would do in this exact scenario is apply an inter-frame (a.k.a. temporal) filter. The specific filter I'd recommend here is the Temporal Cleaner filter available for VirtualDub, using lower-than-default settings. The downside to a temporal filter is the potential blending of motion, a difference kind of blur (blur across time, instead of blue of resolution/detail like the intra-frame filters), however we will avoid or other minimize this kind of damage by only using narrow settings. I halved most settings, and then further lowered a few of them by one more integer, each time testing the results, balancing noise reduction with temporal artifact addition. I did not observe any time-based/temporal blurring, with this exact filter setting applied.

Although I believe this filter is part of the base install of VirtualDub, I have attached the filter (inside a .zip) to this post. You go to FILE > LOAD PROCESSING SETTINGS in VirtualDub to load the video filters, compression settings, and audio settings.

NOTE: If you are using Deemon Video Enhancer (DVE), which supposedly takes advantages of multi-core CPUs, then you'll need this file, as DVE does not come with filters pre-loaded. I've tried to use DVE myself recently, on a recent project, on an AMD Phenom X4 (quad core) system. DVE was slower than VirtualDub, which in turn was slower on that system, due to a 1.8Ghz per-CPU clock, than my 5-year-old AMD 2700+ which sports a higher Ghz single CPU. Maybe it will work better on Intel-based multi-core systems?

That took care of the noise.

Your next issue is the noise patterns found in the overscan area. You want to mask this with black. In VirtualDub, this is accomplished by applying the nul transform filter, and then cropping the image down. Immediately after that filter, apply the resize filter, and check the option that expands the frame back to your original size of 720x576. If this sounds confusing, just load up the VCF filter, attached the bottom, and examine my filters and filter settings.

When you crop down the overscan on interlaced footage, you must use even numbers on the vertical cropping at top and bottom. In general, you never want to use more than 10-12 pixel mask at top/bottom, or 16-20 pixel mask on left/right, as this is close to the viewable image that will be shown on screen. It is never ideal for the mask to be visible, the screen image should bleed to the edge.

If your image is off-center, as this sample seemed to be, then simply mask out the noise areas, plus +2 more beyond that if you're not at your max of 40 pixels (total for both left+right), or 24 pixels (total for both top+bottom). When you apply the resize filter, it will re-center the remaining image.

You NEVER want to resize interlaced video, it introduces a lot of temporal artifacts that get ugly. It can be done, but this is best avoided at all costs. I've not had to resize interlaced video in more than 5 years, it's just not commonly needed.

The overscan in your sample seemed to shift on varying scenes. I assume your compilation might have been made from different tapes. This is what I observed:

Attachment 195

On this exact scene, the overscan needs to be cropped, and then the image re-shifted back to proper center, for its final archiving. Because the overscan shifted several times, you'd want to process those scenes individually, and then re-merge in the editor. My blue arrow in the above image illustrates how the video will be moved left and downwards, once it is re-centered. I had to cut more off the left and bottom, as opposed to the right and top.

I've uploaded the restored video compilation, to show how the temporal cleaner filters help improve quality. The crop adjustment done in this video is adjusted for the very first sample image. I did not cut the clip up and re-crop/re-align every different kind of scene. This will become evident as the clip moves along the timeline, and you can notice rough-edge overscan noise in the frame made by my VirtualDub crops. See sample at http://www.digitalfaq.com/downloads/...n-filtered.rar (154MB file, HuffYUV AVI clip inside of RAR file).

I heard some hiss in the audio, but the sample provided was too short. Audio is harder to work with on a sample basis, I'd need an entire clip, to isolate a noise pattern, and then apply it to the head of the file, to process in Goldwave and/or SoundForge. Audio is easier to work with, however, than video, as it's much faster to process, and the files are much smaller.

This is all assumed to be DVD conversion, streaming video would have added steps to pre-process the video, and make it suitable for H.264 or FLV conversion.


.

tobias 04-13-2009 05:52 PM

Thank you for the advice. I stumbled upon tapes of not as good condition now aswell. Tapes that has material that originate from Super8 (1970s-material). I'll post samples tomorrow, don't have time right now.

What's would you say is next in the workflow then? How would you move on in my shoes after having the filters active in virtualdub? Save to avi (will take quite some time) and then import into a dvd-authoring software and convert to mpeg (which again takes alot of time)?

What dvd authoring software would you recommend? I won't be making any menus, but indexes (chapter) would be needed. Since I have Encore, perhaps I should import the filtered files into premiere to mark the chapters and then export to MPEG2-DVD?

lordsmurf 04-13-2009 06:11 PM

In terms of workflow, you really only have two choices:
  1. Encode from VirtualDub to HuffYUV (or any low-compression/uncompressed codec), and then encode again in an encoder.
  2. Frameserve from VirtualDub (by way of AVI Synth) directly into an encoder accepting of frameserving. For hobbyists, this is TMPGEnc Plus. For professionals like us, we'll yield a higher quality return using either Procoder or MainConcept. (I prefer MainConcept.)
  3. Actually, there's a third option too, discussed further down...
Notes:
  • I would never let an authoring program encode video, it will never be as good as a manual encode to MPEG. Only use authoring software for authoring.
  • Sometimes frameserving from VirtualDub to an MPEG encoder takes longer than simply saving to an intermediary AVI file, then re-encoding to your final MPEG. This is due to multi-core CPU access. Frameserving speeds are something you'd just have to test on your exact system. It can vary.
Authoring: For menu-less authors, look no further than TMPGEnc DVD Author 3 or TMPGEnc Authoring Works 4. TDA3 and TAW4 are pretty much the same, but the newer TAW4 comes with the ability to do basic Blu-Ray authoring, too. The programs are extremely easy to use, simple drag-and-drop interface, very inexpensive software to buy, and chapter marks are simple to use. An optional AC3 encoder is also available, and it encodes to fully-compliant Dolby Digital AC3 (unlike some other tools out there). A guide for TDA is available here: http://www.digitalfaq.com/guides/vid...or-tsunami.htm

Third Option: From the sound of things so far, for this project it seems you may be able to bypass Adobe products entirely. However, if buying a standalone encoder is out of the question, such as MainConcept, then your third option is to encode to uncompressed/low-compression AVI from VirtualDub, import into Adobe Premiere, and then export from the Adobe timeline using the Adobe Media Encoder (based on MainConcept), to a DVD-compliant MPEG-2 setting. Then take that MPEG and author in something like TDA3/TAW4.

tobias 04-13-2009 06:37 PM

Ok, will do it without the frameserving this time around. MainConcept wasn't cheap, 290€ for the MPEG-2 plugin alone, but that's the only package I need, right? So if I buy it, any specific settings I should aim at? (PAL VHS-material, and also some old Super8 material that has been transfered to VHS)

lordsmurf 04-13-2009 06:52 PM

Please use this link -- http://esales.element5.com/product.h...1d6809fe01d97a -- when you buy the MainConcept MPEG-2 Encoder for Reference. With all the copies of MC we help them sell, may as well get a tiny % of the sale, to help support this site we're using.

The Reference "base program" itself is free, download at http://www.mainconcept.com/site/pros...ion-20381.html online. The included codecs come in demo mode, with a watermark, until you purchase the individual codecs. Reference + the MPEG-2 Codec are essentially identical to the older MainConcept MPEG Encoder v1.4/v1.5 from a few years ago, when MainConcept earned its reputation as the best MPEG-2 encoder amongst pros.

While Premiere CS3/CS4 comes with a lot of included MainConcept output options, I've found a number of times that skipping Premiere, and using MainConcept Reference directly, will yield faster encode times. I only use Premiere export when I'm doing an editing project inside of Premiere CS3. I don't use it for conversion, or a number of restoration projects.

There are some special settings to consider, yes. I'll post more on that here in a bit. It's not really "special" as much as just adequately tweaking the bitrate, the number of passes, and other relevant standard quality settings.

tobias 04-14-2009 09:57 AM

Eagerly waiting your recommended encoding settings. I assume these apply to whatever software I chose to use, so.. I have two of the three most important tapes now filtered and ready to be encoded. Have sent the audio to be cleaned up at the studio (using Sadie/Pro Tools) so I am finally starting to see a light at the end of the tunnel. Got in bed at 7:20 in the morning today, the project is starting to tear on my soul :) Thankfully my cold has left the building.

tobias 04-14-2009 12:33 PM

I can give a bit of information about the videos. All are VHS-source, one is originally Digital8 but has been transferred to VHS in -92.
One of the "pure" VHS-sources is of good quality (the one I also posted the compilation from, note I posted the worst clips from it!).

The lengths of the tapes are:
- 2h27min (good qual. tape)
- 1h15min (semi quality tape)
- 2h44min (old material (70's)/low resolution)

Any possibility to fit them on one DVD per tape? I'm guessing a bitrate of ~3500 (VBR) for the better ones? How about the resolution? I've read on these forums that one should set the width to 352 but still keep the height at 576 (PAL)? I haven't really gotten around this fact yet, or it's no fact, just something I remembered me reading sometime.

lordsmurf 04-14-2009 01:05 PM

Quick answer for now...

For single-layer DVD media, you don't want to put more than 2 hours on a disc at 720x576 -- less if the material is shaky hand-held type footage. The bitrate for 2 hours of material is about 5500k VBR 2-pass.

For 720x576 beyond two hours, move to DVD+R DL (or DVD9 press) media.

For more than 2 hours on single layer media, you really need to consider dropping resolution to 352x576 for optimal picture quality. Since the source is VHS or worse, your sources are already closer to 352 than they are 720. There won't be any perceived quality difference, especially not with the high-end hardware and software being used here (some consumer-grade hardware/software appear soft at 352x when it should not).

For those 3-hour videos, yes, a bitrate of around 3500k VBR 2-pass would be correct.

The VHS compilation you posted was by no means awful, it's well within the realm of being restored so nicely that a person would think the VHS tape was made from it, rather than the other way around. So if that was the worst, then those tapes should be fairly pleasant to work with.

tobias 04-14-2009 01:20 PM

Thanks for clearing that up for me, time to start encoding. We'll discuss this deeper later on.

No no, that compilation was of the worst parts from the first tape (the best tape of the batch). The other two couldn't compare to that one.

Sadly there were some rainbow-effects that I didn't manage to remove, but I will have to give then a new dvd from that tape later on, since they require the first batch by tomorrow. I'll send a sample of the bad parts as soon as I have time for it.

lordsmurf 04-15-2009 03:27 PM

Rainbow effects are sometimes the result of temporal chroma errors...

... assuming you've used a good high-end VCR with an embedded TBC, and it was enabled during capture. This TBC would have wiped out in-frame chroma noise, leaving only the ghostly chroma trails, a by-product of certain filters, either in hardware or software.


All times are GMT -5. The time now is 07:25 AM

Site design, images and content © 2002-2024 The Digital FAQ, www.digitalFAQ.com
Forum Software by vBulletin · Copyright © 2024 Jelsoft Enterprises Ltd.