Sorry if i'm doing something wrong in my first post.
Basically i'm trying to restore some footage i got from old VHS tapes. For now i have two different videos to play with and want to improve the image before storing them (in h264/mp4?). I may later try to recapture the tapes (i think i know where they are), for now i have the huge files compressed with lagarith (should have used huffyUV or other?).
I've spent some free time over the last two weeks reading about the subject and i *guess* i need to deinterlace the video, maybe do some color/black&whites correction, maybe cropping and possibly play with noise reduction or sharpness - anyway i accept any suggestions.
My main problem is related to the amount of guides and ideas i read around the web. Read a few such as Capture guide from Doom9, one of the guides in the digital faq, others (e.g.: http://www.engon.de/audio/vhs4_en.htm) but while they are highly interesting in explaining the concepts, i think that most seem outdated regarding the plugins to use. A simple example with deinterlacing, i've checked 100fps.com in addition to the other articles and tried (briefly) some of the suggested plugins. Still the best result i got (at least to my naive perception) was with the virtualdub internal deinterlacer using the Yadif + doubling the frame rate - should probably be using avisynth, right?
Attached are 3 pictures of the videos (two from video1 and one from video2).
What kind of corrections do you suggest and possible plugins to achieve that? Thanks a million!
+ Reply to Thread
Results 1 to 17 of 17
Before record anything buy a TBC to make image captured correct.
Right, both an SVHS with a line TBC as well as a full-frameTBC (although I sometimes choose not to use the second).
In your researches I would think you'd have already come across those recommendations.
Lagarith is fine. It's size is actually a little bit smaller than the one HuffYUV would make. And yes, AviSynth has the best deinterlacers. If for DVD though, you shouldn't deinterlace. For h264, I suppose you should.
Thanks for the suggestions.
You're right manono, I've read a bit about that but haven't really considered buying given the high prices and difficulty in obtaining them (at least here in Portugal). I've borrowed a JVC (not SVHS nor near the prosumer models i saw recommended around) and i've looked at this TBC: CTB-100 (+22£ shipping), but the CTB-100 alone costs 1/3 of the *average* wage here (not the minimum!). Considering that i'm still studying and only have 3 or 4 VHS tapes, my goal was to be more pragmatic and use the available tools to save these.
Is that my cheapest option in Europe/PT regarding TBCs? Are good PAL S-VHS still available? I guess i wont be able to afford either...
Welcome. You certainly have some work cut out for you. But you've made your captures in the preferred manner, to lossless AVI in YUY2 colorspace. I'd prefer Lagarith over Huffyuv, the latter having gone through many permutations to the poiint where a huffyuv plays properly on one computer but not on another. Lagarith is more widely compatible.
Deinterlacing has a quality cost, the cost level depending on how it's done. Avisynth has the best (QTGMC) with Yadif pulling up second. Yes, there's considerable difference between them, and not difficult to see. I'd advise archiving the originals as interlaced. Deinterlacing in software and TV has come a long way, but that will improve in the future. Once you discard an interlaced original, it's gone forever; future improvements in technology couldn't be exploited.
There are many who upsample standard definition captures with the idea that they will "look like" HD. They won't. They'll look like upsampled standard definition. Various upsampling techniques have varying results.
Why are you cropping? DVD and BD have frame size requirements, referred to earlier. Cropping is usually done to center an image within borders and to mask bottom-border head switching noise (which one can see in your images).
VirtualDub. One problem shown in your images is the top-border flagging. This is due to playback problems. It's been suggested that you buy a tbc. That's an incomplete suggestion. There are two basic types of tbc: line-level and frame-level. For VHS media, a frame tbc is useful for overriding copy protection, but most VHS playback problems are solved with a line-level tbc. A frame tbc would have no effect on the top border flagging nor on many other typical tape playback problems. There are times when tape is so mauled that both tbc's are needed, but a line tbc is more essential.
Digital transfer requires a perfectly timed signal. That's not what you get with VHS. Tape wobbles in the playing path, the playback speed varies even if slightly, so lines in the frame aren't always transmitted exactly "on time". The result is often flagging, as well as lines and edges that wiggle somewhat. Some tape players have built-in line and frame sync tbc's, but these don't always do the work required -- some of them can make damaged tape look worse. Unfortunately, shop-grade line tbc's are prohibitively expensive (I don't know why), while decent frame-level tbc's start at a lower price point. A popular way to achieve a line-level tbc is to use a DVD recorder as a pass-thru device. Rather than record directly to DVD with these machines (which is NOT a good solution), the recorder is used as a "play-through" device between VCR and capture device. Many DVD recorders will work as tbc pass-thru, but not all of them. Today's DVD recorders are substandard relative to older models of the 2002-2005 vintage. The most numerous of these are from Panasonic and Toshiba; there are other brands, but those "others" that can be used for pass-thru are touigh to find. There are also some high-end camcorders that can be used for pass-thru.
While a pass-thru device is no competition for pro gear, pro stuff requires associated equipment. The consumer-level DVD recorders that most people use are affordable and easier to operate: you just connect them and they do their work. There are a number of posts in this forum that discuss these methods. Some of those posts get downright techy, others are more simple. I recently posted several examples of DVD pass-thru in use, and one of them demonstrates the variable results I've had with a couple of pass-thru machines. The examples are here, and they particualrly discuss flagging issues: http://forum.videohelp.com/threads/331681-s-video-artifacts?p=2141386&viewfull=1#post2141386.
There are other threads that discuss pass-thru, notably this one with lots of charts and cryptic discussion:
As for color correction and noise reduction, well.....that's a huge ball of wax. A hairball, really. You'll find widely differing opinions. VHS tape has unique noise and color problems. Tape noise itself is a subject all its own, for which temporal filters are usually recommended (and sometimes not, depending on the character of the noise). There are also sharpeners (almost always overdone, which looks horrible) and line and edge conditioners. Most of the best noise reduction filters are slow performers, but good work takes time. QTGMC is a good performer in a general ways, and can be used on progressive and interlaced material (and you can always reinterlace, which is not difficult). I note some fuzzy edges in your posted images; QTGMC and MCTemporalDenoise are adept at smoothing those, but you can use many others. The best way to get a handle on all these plugins is to look over posts in the Restoration forum and note the filters and methods recommended by experienced users. You'll find that the same group of plugins occur frequently. Most filters work best on video that is deinterlaced (and reinterlaced if necessary) or whose interlaced fields have been separated in Avisynth and reweaved after processing.
I would recommend some ready plugins, but others are bound to chime in with more debate. The one VirtualDub filter I most often use with VHS is NeatVideo, which is not free but is not that expensive, is worth its weight in gold, and is IMO essential for many rotten tapes because it often addresses some issues that are difficult to solve with other plugins. But NeatVideo can't be applied to everything that comes down the pike, and many people either don't understand how to use it properly or won't bother. VirtualDub has a built-in temporal smoother that's quite good at lower settings.
Filters tend to be rather specific. There are not many one-size-fits all filters, and the more filters you use the more odd your video looks. It's difficult to recommend specific filters based on still images, as most filters are designed to work with an image stream wherein what happens in one frame depends on what happens in the frames around it. So I would say that browsing the Restoration area will yield many ideas, and submitting a chunk of your own problem video would be a good way to get started.
Color correction is also a huge field. Most beginners attempt to judge the corrections by eyeball alone. That's an exercise in futility, although the human eye can certainly know when something isn't quite right -- but the eye doesn't always know exactly why things don't seem correct. So you should learn to work with histograms and various 'scopes, all of which are free. Many learn to use these tools and get insight into color theory through free internet tutorials. These little lessons are quite informative, even if they deal with the likes of Photoshop Pro, digital cameras, and still images. However these are useful because the principles of correcting images and using the tools are the same for video and still photos alike. One corrects saturation or a green color cast in photos and video using the same principles and techniques.
The bottom photo of the woman has saturation problems; it might be possible to clean up the cyan horizontal stripe near the frame's bottom, but we'd have to see other frames to be certain. The picture with the boy looks undersaturated because its basic gamma level is too high; the tape was damaged and possibly bouncing in the tape path, causing the rip and jitter effects seen. In the middle photo the bright highlights are clipped (which means they are destroyed) during capture or when converting colorspace in VirtualDub; there's aliasing and red chroma shift.
Avisynth and VirtualDib have useful tools to analyze and check your corrections. In VirtualDub there are the ColorMill and gradation curve color filters, and RGB histograms and vectorscopes. There are similar tools in Avisynth.
I'd suggest that you post a sample video that are problems for you. Then things can get more specific.
Last edited by sanlyn; 24th Jan 2013 at 21:50.Our inventions are wont to be pretty toys, which distract our attention from serious things. They are but improved means to an unimproved end. -- Henry David Thoreau
What a huge answer, thanks sanlyn!!
Don't even know where to start but will try to reply to some of your points briefly.
Yes, i've captured using vdub to lagarith YUY2 since it was described in a recent topic.
I wasn't really clear with the h264 part. What i really wanted to say is that my objective is to save the footage to a digital format. In addition, make copies in a digital, compact (and so, not lossless) format such as h264 (just picked this because it seems to be selected to release movies and such, achieving a good quality / size ratio). This would allow me to easily access and share them in a usb / external drive or play in any recent TV (or say computer, tablet) to some relatives. As for achieving purposes, i'm aware that lossless files should be kept.
Hope i'm doing a decent choice. I could convert them to DVD but i could also burn the mp4 (or others) to DVDs. Are there many advantages in authoring a DVD and using mpeg2?
So the CTB-100 falls in the frame-level TBC group? Didn't know about the usage of DVD recorders to that end, is it preferable over external TBCs such as the CTB-100 or they are doing different jobs? I've searched ebay for Toshiba D-KR2 and Panasonic DMR-ES20 but sadly (mostly) ship only to US. Probably i would be better served with the CTB? I'm in Europe and my recordings are all PAL indeed, would that influence the pass-thru with recorders from the US?
I see the most important part is capture and i should probably start saving a bit of cash to improve that. As for the restoring, it is indeed complex and i guess there are no easy recipes, it is all dependent on the source, personal taste and tools available. Indeed i will probably pick a short clip and do experiments based on your suggestions and knowledge gathered around here. Just to justify the deinterlacing option, i was considering doing it since it is one of the first steps described in the doom9 guide. It suggested to deinterlace (when possible without loosing quality or to encode to other formats) except if the idea was to encode to DVD - with i was not considering. Anyway i'm really inexperienced in the field and trying to (hopefully) improve my knowledge/results before screw up things
I would suggest archiving the original lossless captures. You can always go back to them to make other versions as technology improves. A Lagarith YUY2 capture is a "digital" capture. A lossless or nearly lossless h264 archive would be the next best thing.
Not everyone rushes out to keep pace with the latest digital players. Many of my relatives bought standard DVD players years ago and are likely to keep them forever. Older players can't handle newer h264 or BD, so if you plan on making gifts of those videos and the folks use old players you're kinda stuck with good ol' MPEG2 DVD. That will change with time, which is why it's a good idea to retain the original captures as-is.
The CBT-100 won't address the problems shown in your images. It's not necessary to use the same DVD recorders as in my examples. In fact, the optical drive in my old Toshiba won't recognize many discs any more (although my newer ones work work well, no problems). There are many models of those players from Panasonic and Toshiba. They were quite popular here and in PAL country -- so much so that many owners still use them. You might come across some selling cheaply because the optical drive died -- you don't need the optical drive for pass-thru. You'll find other discussions in this forum about pass-thru devices that mention several alternatives.
I'm not a fan of deinterlacing originals, for the reasons stated. With some digital disc formats, interlace is the standard. Remember that PC playback is not the same as playback thru a home a/v system that deinterlaces during play. Many people fight city hall and bypass the standards. I don't believe in that, and I've seen precipitous results depending on what's playing the rulebreakers. But that's up to you. There will come a time when interlace disappears entirely. Not yet, though.
You might notice that if your old tapes have been played many times they are not wound smoothly on the reels. Sometimes repacking tape can get smoother playback. You repack by fast-forwarding the tape (without playing and without pause) to the end, then fast-rewind (without playing and without pause) to the beginning. You might have to repeat this process. The object is to see that the tape is more smoothly wound on the reels with fewer bumps and ridges.
Last edited by sanlyn; 25th Jan 2013 at 13:07.Our inventions are wont to be pretty toys, which distract our attention from serious things. They are but improved means to an unimproved end. -- Henry David Thoreau
Yup i plan to achieve the lossless captures, not sure what the best format is but probably just as they are captured with lagarith? Or after some processing?
I see your point about the DVD, i think i even have cases of older relatives without DVD players, the main idea of having a h264 version was to have an easily accessible file in my laptop or in a table or something. Anyway, with the capture and recover part done i guess encoding to h264 or dvd is not a problem, i could do both. As for interlaced, i guess i thought progressive was the general choice but i see it really depends on some factors. So i should really try to get a decent DVD recorder (or player?) from Panasonic or Toshiba around here, maybe i can find some in my city. Any models i should avoid or key specs i should look for that are needed / essential for pass-thru?
You captured correctly the first time, I believe: to AVI with lossless Lagarith compression in YUY2 colorspace. Eventually you'll encode into YV12 (the standard) but YUY2 is "closest" to the way your tapes are playing. From your original you can go wherever you want. Once you've made your corrections as needed in AVI, the field is open for encoding/authoring. If you author to optical disc, check the specs for the disc format you want: h264 is for BluRay and/or AVCHD, vanilla MPEG2 is for standard DVD. Since apparantly your need for SD DVD won't be as frequent, then h264 should be your regualar route. As long as you have your saved corrections, you can go several ways if you need to.
Almost all of my VCR's and some of the older DVD gear came from eBay, but for me that was the most convenient source. There are precious few DVD recorders on the current market from either of those brands (or anyone else), and newer models are inferior in all respects to the 2002-2005 era. You'll find Panasonics in ES10-, ES15, ES20 models, Toshibas as the RD-K2, -3, -4, -5 and the RD-XS series The old RD-K2 isn't that great if you want it as a recorder, but its tbc is OK -- and Toshiba's RD-"XS" series are valued models that are hard to find, and pricey. The "K model numbers vary in some countries, they just have the same basic letters and numbers in different order. Mind that Panasonic's pass-thru tbc is available on the Line 1 input only.
In PAL country you have more new models from Panny and Toshiba available. Pitiful old NTSC land has only 1 recorder brand sold today ("Magnavox"), an OK deal but nothing like the older ones. Not all DVD brands can be used for pass-thru (Pioneer and SONY, for example).
In any case, post a brief sample when you can. No one here can give detailed advice without the genuine article at hand.Our inventions are wont to be pretty toys, which distract our attention from serious things. They are but improved means to an unimproved end. -- Henry David Thoreau
What are the major advantages of archiving in YV12 instead of just YUY? I will look at that later, my main priority at the moment should be to recapture using something as pass-thru. Saw someone selling a Pioneer DVR-3100 in my area but you refereed to Pioneer as unable to do pass-thru. Right? Must keep looking for some good deals around here. The other two that i found near me were the Samsung DVD-R129 and Sanyo DVD 7830RW, guess i should wait or look better for Panasonic and Toshiba ones.
Is it useful to post a brief sample of what i have atm or should i just wait until i get better captures? Any special part i should post? Movement, lots of colors or whites, how many seconds? Just asking because currently 20secs results in almost 200mb in the current form.
Last edited by PandMonium; 26th Jan 2013 at 06:36.
VirtualDub or other viewing apps), then it is saved in the YUV form in which it was captured, whether it's YV12 or YUY2. YUY2 is preferred for analog capture, it stores more chroma data than YV12.
20 seconds of PAL AVI shouldn't quite hit the 200mb mark. But 8 secs can tell much and should fall within the 100mb limit here. When you open YUV video directly in VirtualDub, the YUV has to be converted to RGB for display. If you run RGB filters or otherwise save the clip in processing mode, RGB will double the YUV's original size. Do it this way: Open the original YUV capture in VirtualDub, then edit for the sample you want. Before clicking "Save AVI...", Go to "Video...." and click "Direct stream copy". This will save your cut in the original YUV colorspace instead of RGB.
A cut from the scene in your middle (outdoor) image would work, as it contains many straight and angular lines, and a small amount of slow camera pan would suffice -- avoid really shaky camera jiggle or fast pans and zooms. A scene from your bottom photo also has straight lines and edges, and movement of the hands and arms or moderate horizontal pan would suffice. The top image has motion, but not many straight geometric forms to assess line timing output (nothing personal against the handsome boy, but neither he nor the plants have a lot of "straight or angular").
Sorry for the delay but was busy during the weekend.
Thanks again for all the valuable tips and here is a sample: https://dl.dropbox.com/u/4480058/scene.avi (a bit shaky but hope it is ok).
What do you think should be my main concerns, am i too far with my original ideas?
Nice color. Most of the time.
Some of your worst enemies (common to most of us):
1. Autogain (aka auto-exposure).
3. Trusting that the camera is smarter than you can be.
4. Belief than you are steady enough to live without a tripod or shoulder brace.
5. Capturing without considering valid luma and chroma levels: crushed darks, blown-out highlights.
6. Capturing old tape without line-level TBC.
All of these are up for discussion, of course .
I made an effort, attached below. Far from a perfect fix. I used Avisynth to correct the bright levels somewhat, although the camera and the capture combined to blow away highlights.
frame 124 (original)
frame 124 (levels adjustment -- subject to endless fiddling, and shown here a bit darker than it displays to emphasize the tamed brights):
Some of it can't be fixed. The frame motion is a little more steady, but of course that's no substitute for a tbc. The tape plays as if it were literally bouncing, rattling, and in jerk-motion thru the tape path -- which is really what was happening. The player itself outputs a decent image, but few players can handle damaged analog media with perfect timing. Even a "pro" player needs help from tbc circuits.
Last edited by sanlyn; 30th Jan 2013 at 02:54.
Nice to know that colors are ok. I also thought it was probably too bright since there are lots of whites and possibly loss of details (like the sky and floor). About the mistakes, you're right. I think the person who filmed it decades ago wasn't really aware of any of that, just used everything auto/default and recorded (which imo is the usual proceeding).
Your captured frame looks good, it was deinterlaced by the player or what? My ignorant question is about what improved the lines in the movie, like the handrail in the stairs and its shadow. Also, considering the 5th point about "capturing without considering valid luma and chroma levels: crushed darks, blown-out highlights", is this something that can be improved with new captures? Where did i went wrong or miss?
# ---- Bring luma and chroma into valid range RGB 16-240). ----- # ---- Some parts of the sequence defy good correction. ----- ColorYUV(cont_y=-40,gain_y=-15,gamma_y=-15,off_y=-12) ColorYUV(cont_v=-15,cont_u=-15) ChromaShift(L=-4) Crop(4,6,-12,-10).AddBorders(8,8,8,8) # ---- Deinterlace, discard the odd fields, clean up a bit ----- QTGMC(preset="slow") SelectEven() TTempSMooth() ConvertToYV12(interlaced=false) mergechroma(aWarpSharp(depth=20.0, thresh=0.75, blurlevel=3, cm=1)) # ---- make 2 copies of results, then re-interlace ----- in1=last in1a=in1 in1b=in1 in1c=Interleave(in1a,in1b) In1c AssumeTFF().SeparateFields().SelectEvery(4,0,3).Weave() return last
Below are two RGB histograms. You can use others, but I often convert to BT601 matrix in RGB to see how video will display on TV, as opposed to how it's stored in YUV. The graph on the left is the original frame 124 with black borders removed. Luma and colors are shifted to the right (yes, it's too bright). At this point luma (the white area) was within the 16-240 range indicated by the yellow lines at each side, but colors were smashed against the right-hand side -- indicates clipping of brights. I used Avisynth filters to bring all those values into line (histogram on the right). You can see that the image lacks a little "snap" because luma had to be contracted to rein in chroma, and the hard peak of luma at the right shows that values were clipped during capture and can't be fully recovered.
Because these values changed a dozen times during the clip, I adjusted for worst-case and best-case averaging of values. The camera's autogain was having fits, so no two histograms from the frames look alike. When it came to the flower closeup, the highlights were destroyed and darker areas simply had to be sacrificed for the main subject; the contrast range was too wide for complete correction, even for PC playback.
Last edited by sanlyn; 30th Jan 2013 at 10:29.
That's interesting sanlyn, especially the RGB histograms. I see there is always a tradeoff, to fit the colors you have to contract luma, degrading something to improve other parts (and get a global improvement). With the flower closeup you were saying that the source was like "too damaged" to be recovered (no information left) or something related with the fixes from the previous scene that would affect negatively the close up? I guess the videos normally are split in scenes where things change a lot (like bright outside followed by an inside, darker part?) in order to properly apply different settings to each scene, right?
I'm really curious about the results using a line tbc but as you said they are hard to find. Being pragmatic, i think i should capture the tapes now while i have a working setup and later if i find one i could recapture them. What's your opinion on this?
Meanwhile i've tracked a Sony RDR-GX210 near me. Just googled a bit and saw some people trying (and sometimes complaining) on using it as a pass-through, probably is not a good choice too and should keep looking. Right? :/
The contrast range in the flower scene is too wide to be recorded on tape, or even on movie film. Setting a manual exposure might have helped (but only somewhat, not completely). If you look at the histogram again, especially the one on the right-hand side of the image posted, you can see a hard, high peak at the right; the peaks indicate clipping. The bright areas were too bright to be recorded, so at a certain point in the spectrum the luma and colors were simply "cut off" -- this is what clipping means. If you play the video in VirtualDub with a histogram showing in the "results" window, you'll see how often the histogram changes--the brightest yellow could not be recorded, so the "color" changed from yellow to nearly white. Making the brights "darker" would not reveal more detail.