Auto color a la Photoshop

[omnimoeish]

Hi all, this is my first post. I have about 20 home VHS tapes I'm hoping to archive (just on the computer, not to DVD). I just bought a Canopus ADVC-300 and I am borrowing a JVC SVHS HR 4600U (beggars cant be choosers) and I'm using my decently calibrated 1 year old Samsung 46" HDTV as my computer monitor and TV .

Anyway, I use Adobe Photoshop CS5 and its remarkable how the auto color can fix picture color. Sure I dabble in manual color adjustment but I find the auto color works pretty well 90% of the time and when I'm trying to fix tens of thousands of pictures, I just don't want to mess around with them all if I don't have to.

I was just wondering if there's any program that's based on similar algorithms that can fix all the frames in a video? I figure with a 6 core processor and the fact that these frames are 2/3rds of a megapixel, it probably wouldn't take too long for the processor to do it, provided there was software available. If it doesn't exist, hopefully some of the programmers can figure out how to make a Virtualdub filter for it or something because that would ROCK.

[redwudz]

If they all have the exact same color errors, you can just tune one of the filters like Color Mill 2 or similar and process each video, leaving VD running in between. Or note down the filter settings. But generally, unless all the videos were shot with the wrong color setting, florescent instead of incandescent or natural light, I doubt you can do automatic correction. And if you are going to filter, there may be other non-color filters that would help with VHS video.

Also, a bit more advanced, would be a AVISynth script so you would have full control over the filtering and have it easily repeatable. But someone else would have to advise you on that.

And welcome to our forums.

[omnimoeish]

Yeah, I think for the most part they're ok, the main one I'm worried about is my parents wedding video which will be 32 years old Jan 18th. It's got all kind of bizarreness going on. The Canopus allows you to fix some stuff in analog before converting to DV, which is nice. It seems like there are certain segments with certain issues so it takes some patience to stop and go and adjust and go back and recapture with the new settings. Even after all that, it's still pretty limited. Plus some segments have double hue problems (ie one side is green and one side is purple) so I'm scratching my head on how to fix that. I'll see if I can try your suggestions for further refinement using lossless transfer I suppose. Avisynth scares me though. I see that there is some stuff I can do in Adobe Premiere although that's going to require a lot of patience too. Oh well, that's half the fun.

[omnimoeish]

I did about 10 frames just to see what it would look like, this one is nice to fix since it's all blue.

[jagabo]

You're looking for an auto white balance filter (and maybe auto gain too). Beware though that automatic filters like that can cause colors to wander as the picture content changes. And you may want to crop away the black borders first because they can confuse the automatic adjustments.

Your first image after auto white balance and auto gain with AviSynth's ColorYUV():

[Gavino]

If you lose your fear of Avisynth, try
ColorYUV(autowhite=true, autogain=true)

As jagabo says, an automatic filter like this is not a wonder cure, but it should get you something acceptable as a starting point for further manual tweaking.

[jagabo]

If you find the colors and levels wander around too much you can find a frame where they worked well and analyze what the adjustments where made automatically. Then try using those fixed adjustments for the rest of the video. For example, your sample image had the U channel pulled down by about 24 units:



It's not shown here, but the V channel was pulled up by a similar amount, and the contrast (Y channel) was stretched.

[Cornucopia]

That's certainly looking better, but note that it looks like you also have timebase errors that need fixing, and that's best done with an external TBC box (in the chain between the VCR and the ADVC300). It nigh impossible to fix that after the video has been digitized...

Scott

[jagabo]

it looks like you also have timebase errors

That probably is the case but... The camera was moving sideways during that frame and the interlaced frame was resized vertically (notice it's 720x540) without accounting for the fact that it was interlaced. That's the main cause of the wavy verticals.

[sanlyn]

Happy to welcome you, omnimoeish. I think you can see you've already sparked some debate here. We love it.

I use Adobe Photoshop CS5 and its remarkable how the auto color can fix picture color. Sure I dabble in manual color adjustment but I find the auto color works pretty well 90% of the time and when I'm trying to fix tens of thousands of pictures

I use Photoshop as well, and sometimes "Auto" can get you somewhere in the neighborhood of correction. But that's not the same thing as adjusting white, gray, and black levels - look into those areas and you'll see how often Auto falls short.

Viewed with a histogram, your original capture shows more problems than you can shake a stick at. It's not just color balance. There's severe chroma bleeding, chroma noise, magnetic layer damage and fading, etc. Much of this could be repaired in Avisynth, but before converting to DV. A simple Blue gain adjustment with ColorYUV would have helped immensely with Blue and red before going to RGB for more precise corrections. The histogram shows overextended blue, fading green, and crippled red. Keep in mind that AutoColor is also reading the black borders and border stains.

[Attachment 10399 - Click to enlarge]


A histogram of the filtered image shows blue is still a problem, but not the only one. As for the chroma noise, bleed, etc., there are Avisynth routines that can help alleviate much of it - though this video is in bad shape, it could look much better worked in AVI. I've used scripts that fixed some similar border stains. Overall, the photo is still too blue. There are no blacks, grays, or clean whites, and skin tones are off. I didn't try fixing any of this: color damage in compressed DV is almost impossible to repair.

[Attachment 10400 - Click to enlarge]


One thing you can do with AVi captures is to crop the borders and stains so Photoshop can read the image more accurately. In your fix (above) you'll see that Auto Color made a mess of the left border. The histogram looks a bit different with the borders gone.

[Attachment 10401 - Click to enlarge]


Crop the borders, replace the old ones with black borders in Avisynth, and center the image in the frame. There are plugins that can possibly help a bunch to clean those borders and save some image real estate for posterity. A curves filter could do a lot of correcting here.

[Attachment 10402 - Click to enlarge]

I was just wondering if there's any program that's based on similar algorithms that can fix all the frames in a video? I figure with a 6 core processor and the fact that these frames are 2/3rds of a megapixel, it probably wouldn't take too long for the processor to do it, provided there was software available.

I'm not sure what you mean by "similar algorithms". Most advanced users wouldn't touch "Auto". With trouble or damaged video, they work with histograms, pixel samplers, and level and curve controls. Photoshop has levels and curves that use different algorithms; their settings can be saved and imported into other Adobe apps like Premiere Pro and After Effects. In fact After Effects is likely more useable for working with damaged video. Neither are cheap or easy to learn, but they're way ahead of "Auto".

If it doesn't exist, hopefully some of the programmers can figure out how to make a Virtualdub filter for it or something because that would ROCK.

The VirtualDub gradation curve rocks. Works like curves in Photoshop, Premiere, etc., and can import Adobe's curve setting files. VirtualDub has a specialized ColorTools histogram, and there's a free CSamp pixel reader .

the main one I'm worried about is my parents wedding video which will be 32 years old Jan 18th. It's got all kind of bizarreness going on. The Canopus allows you to fix some stuff in analog before converting to DV, which is nice.

Are you converting to DV, or recording to DV? If recording directly to lossy DV compression, it's not so nice. Cleaning up compressed DV is really difficult. Recording to AVI with lossless huffyuv or lagarith in YV12 or YUY2 gives you a fighting chance. After chroma noise, tape noise, compression artifacts, color stains, etc., get digitized into compressed DV you're in trouble.

It seems like there are certain segments with certain issues so it takes some patience to stop and go and adjust and go back and recapture with the new settings. Even after all that, it's still pretty limited. Plus some segments have double hue problems (ie one side is green and one side is purple) so I'm scratching my head on how to fix that.

Record to AVI and avoid all that.

I'm using my decently calibrated 1 year old Samsung 46" HDTV as my computer monitor and TV

Many people use a PC->TV setup. It's great for testing the final output video. Most photogs and videophiles would use a PC monitor, which can be calibrated to closer color and luma standards than most TV's. Calibration for PC monitors is done with a colorimeter and specialized software, such as this kit - not the best and biggest, but the most popular recommendation for hobbyists and pros alike (and many sites sell the iOne-2 far cheaper than shown here): XRite EyeOne Display 2 .
This site shows how it's done: http://www.tftcentral.co.uk/reviews/eye_one_display2.htm
People mean different things when they say their TV is calibrated. Here's how you do it with a colorimeter: http://www.curtpalme.com/forum/viewtopic.php?t=10457

[omnimoeish]

Thanks for the responses, I'm glad it looks like there are some options that do what I need. The reason I was hoping for something relatively automatic is that (I have some other frames I could show you from the clip if that would help, but) there's certain segments that are very blue, some are very red, some green, it's got all kinds of stuff going on.

I just talked to my grandpa and he says he has about 40 V-cord videos from the early 70's of my mom's childhood. I'll bet those are gonna need some color correction.

Are you converting to DV, or recording to DV? If recording directly to lossy DV compression, it's not so nice. Cleaning up compressed DV is really difficult. Recording to AVI with lossless huffyuv or lagarith in YV12 or YUY2 gives you a fighting chance. After chroma noise, tape noise, compression artifacts, color stains, etc., get digitized into compressed DV you're in trouble.

As far as I know, Canopus ADVC-300 only converts to DV. It's connected by a 6 pin Firewire and I'm "capturing" in WinDV. I know it's kind of a noob compression format but I had some frustrating audio sync issues before and DV fixes this issue. I used to have an ATI All in Wonder X800 but it died.

The 300 model Canopus has the built in (I believe linear) TBC (probably not all that amazing of one) and has analog adjustment controls including brightness, contrast, saturation, hue, and sharpness. It also has 2D and 3D noise filtering, 3D Y/C seperation, and some other stuff that I can't figure out what it does like Black expansion, white expansion, white step, and edge adjustment (vertical and horizontal).

I suppose I'll get into avisynth because it sounds like that's what I need to do. Does it still work pretty well with DV? Anyone used enosoft or DVCorrrector10 to any success? I haven't gotten too into them yet to see if they're worth it.

[jagabo]

Capturing VHS with your ADVC-300 is fine. Just be sure the proc amp isn't set so far out you get blown out brights or darks. If you're going to use software noise filtering you should disable the ADVC-300 primitive noise filters.

Most commercial editor have manual and auto contrast and white balance (and much more) adjustments. I doubt you'll be going to the extents that sanlyn does with color corrections. You don' have to go to AviSynth for simple adjustments like that.

With all helical scan tape formats horizontal jitter is a problem:

https://forum.videohelp.com/threads/306272-Computer-video-capture-vs-vcr-to-dvd-combo?p...=1#post1882662
https://forum.videohelp.com/threads/230650-Confused-Why-a-VCR-with-TBC-if-separate-TBC-...=1#post1346821

That can only be fixed by a line time base corrector. That's usually found in better S-VHS decks. You can also find them in a few DVD recorders like the Panasonic ES15 (the ltbc works in passthough mode so you don't need to record onto disc). Most people don't want to spend US$200 to US$300 for an S-VHS deck (and they're only avialable used) but ~US$75 for a used ES15 may be within your budget.

[Cornucopia]

The built-in TBC in the ADVC-300 is fine for what it does, but I still would REALLY recommend a full-frame TBC (and the line TBC that jagabo mentions above is also important)...

You mentioned V-cord tapes: you won't find ANY decent used consumer playback devices for those (maybe/possibly/rarely on ebay). I would strongly suggest you leave those to a decent professional transfer house.

Even with "automatic" color adjustments, you probably won't be able to do it globally. Works best when you can keyframe the various settings.

Scott

[sanlyn]

Most commercial editor have manual and auto contrast and white balance (and much more) adjustments. I doubt you'll be going to the extents that sanlyn does with color corrections. You don' have to go to AviSynth for simple adjustments like that.

That's generally true. With badly damaged/discolored video that has great personal value, I'd make a special effort. For most video in decent shape it would hardly be worth the trouble. I've often made some decent basic corrections with VirtualDub's simple capture filters.

[omnimoeish]

The built-in TBC in the ADVC-300 is fine for what it does, but I still would REALLY recommend a full-frame TBC (and the line TBC that jagabo mentions above is also important)...

You mentioned V-cord tapes: you won't find ANY decent used consumer playback devices for those (maybe/possibly/rarely on ebay). I would strongly suggest you leave those to a decent professional transfer house.

Even with "automatic" color adjustments, you probably won't be able to do it globally. Works best when you can keyframe the various settings.

Scott

He has a V-cord player apparently, although he doubts it will work without belt replacements. I'm open to professionals doing the V-cord. I don't know how brittle the tape is at this point, or what kind of restoration they'll need and he thinks it'll need a good TBC. What places do you guys recommend that can work on these?

[omnimoeish]

Most commercial editor have manual and auto contrast and white balance (and much more) adjustments. I doubt you'll be going to the extents that sanlyn does with color corrections. You don' have to go to AviSynth for simple adjustments like that.

That's generally true. With badly damaged/discolored video that has great personal value, I'd make a special effort. For most video in decent shape it would hardly be worth the trouble. I've often made some decent basic corrections with VirtualDub's simple capture filters.

If you get off on this sort of thing, it's a video restorers playground. I can send you the DV file I have. It's only about 7 minutes.

[sanlyn]

Nice offer, omnimoeish. Only problem: to do it up right, you'd work with properly capturing the original video (using different hardware) to lossless huffyuv or lagarith AVI. Once damaged video gets compressed into lossy DV, any improvement would be marginal at best. And the effort would take weeks, not hours - even for 7 minutes. But you're right: anyone with more time, hardware and software would have a field day.

[omnimoeish]

I am sure you're correct that there is some loss, but I have yet to see any specific side by side demonstration. Can you show an example of the loss in compression to DV compared to a lossless? The reason I ask is that I'm trying to decide if it's worth sending this tape away to someone who can do it right. Also, how do you do analog editing? Do you have any side by side comparisons demonstrating the advantages of analog color correction over digital color correction?

[sanlyn]

I am sure you're correct that there is some loss, but I have yet to see any specific side by side demonstration. Can you show an example of the loss in compression to DV compared to a lossless? The reason I ask is that I'm trying to decide if it's worth sending this tape away to someone who can do it right. Also, how do you do analog editing? Do you have any side by side comparisons demonstrating the advantages of analog color correction over digital color correction?

What you do with the source depends on how much it means to you. There is no "loss in compression to DV" compared to a lossless format. You don't :lose compression", you lose video data. Most of the damage in recording to compressed format, however, is that bad color, high noise levels (which are exacerbated by lossy compression), and other disturbances become more indelibly imbedded in the resulting capture.

The last time I recorded a piece of bad video to a lossy compression format was about 2003. So I have no examples on hand - because that experience and the discovery of this forum and Doom9 a few months later were all I needed to convince me I'd never capture damaged video that way again.

This forum has countless examples of poor capture procedures and ways to avoid and/or repair them. In this thread there exists only one bad frame to work with; not much can be done with one frame except to play with color a bit, but some of that color damage could have been avoided and made easier to improve later.

Still on my hard drive I have one brief example of a really stupid capture at a small frame size and a bitrate barely above the minimum bitrate allowed for DVD (2500 kbps, which is insanely low for damaged VHS tape). The example is a piece of original unprocessed source MPEG2. Besides obvious problems, color fading, horrible motion control, detail loss, more detail loss with a small frame size, blown-out highlights and border stains, its left channel audio was not recorded (and this is an opera, for godssake!). The original tape was discarded (OMG!), so there's no chance for a better capture.
original (3.4 MB): http://dc181.4shared.com/download/7SlcvoJF/A_A01a9_original.mpg

The entire video was converted to huffyuv AVI in a YUY2 colorspace. Initial correction efforts were in Avisynth. More followed in VirtualDub, and color was tweaked in RGB. While the result does look "better", it's an example of irreparable loss. The original tape, even with damage, undoubtedly looked better than the capture. This filtered MPEG of the same clip shows that after much learning, twiddling and experimentation, there's still a lot of work to do:
filtered (5.5 MB): http://dc439.4shared.com/download/ltlXSSr7/A01a9_DVD_v1.mpg

The more-or-less cleaned version received more tweaking and became part of Version 1. There was plenty of help, guidance and debate in the forum for 14 months. I learned much, enough to make me feel that an effort for "Version 2" would yield better results in 2012. What's so depressing is seeing how someone can make a mess out of something that was already a mess to begin with, but that could have looked and sounded much better.

[omnimoeish]

There is no loss in compression to DV? Got it. J/k I didn't say "loss of compression". The loss I was referring to was indeed quality. Anyway, if I understand correctly, it sounds like there is little to be done in analog as far as color correction as you cite your correcting steps were done in digital anyway (by way of huffyuv of course). So it sounds like the main problem I have is the initial DV compression color loss. I came up with a possible way to test the color loss. I used a high quality h.264 source (may have had color, I don't know). I then converted about a minute to huffyuv YUY2 leaving at same resolution, as well as NSTC DV 25 using Virtualdub. They were both resized to 720x480 (because that's the only resolution DV works in). Here are the pictures. It looks like indeed there is saturation loss. I do however have the saturation up a titch from normal on my proc amp. All in all, I guess I can live with it.

The top one is Huffyuv, the bottom is DV. I am not sure why the snapshot from VLC is not 720x480 when the video is.

Edit, I added PAL DV on the bottom.

Edit 2:I don't know if this is even a fair comparison. Anyone know if png is lossy? Anyway, I can definitely see a slight saturation loss but that's because I'm REALLY looking for it and I'm pretty good at spotting that stuff in general, but I imagine a good portion of the population couldn't tell.

[jagabo]

DV does give a tiny bit of macroblocking from compression. And with PAL DV you will get a slight loss of vertical color resolution (4:2:2 color to 4:2:0 color). Of course, since you're going to DVD you'll eventually lose that color resolution anyway. I think sanlyn is overstating the problem though.

DV has the same gamut (luma and chroma range) as raw YUV (and hence, losslessly compressed YUV).

[Cornucopia]

Yeah, but that's not the best example, as (you say yourself) you're using an h.264 as the source. "High quality" or no, that kind of source is already compromised (particularly colorspace - it's likely 4:2:0!).

Scott

[sanlyn]

I'm referring to the problems of damaged, bruised, crinkled, aged, deteriorated, heat-warped, stained, screwed-up tape sources that are damaged even more by dumb, ill-considered capture procedures. A video source that's in good shape to begin with can take a great deal more abuse.

However, if you don't see the difference between a poor capture from a damaged source -vs- a decent h.264 video, you don't need Avisynth, VirtualDub, or other processing software. All you need is a DVD recorder.

[jagabo]

Anyone know if png is lossy?

It is.

PNG is not lossy if it's given 24 bit RGB and told to compress as RGB and decompress as RGB.

http://en.wikipedia.org/wiki/Portable_Network_Graphics

Portable Network Graphics (PNG play /ˈpɪŋ/[2]) is a bitmapped image format that employs lossless data compression.

But video is in YUV so you will get some losses on conversion from YUV to RGB. Of course, you always get that when you display video because RGB is what shows up on the screen.

Anyway, I can definitely see a slight saturation loss but that's because I'm REALLY looking for it and I'm pretty good at spotting that stuff in general, but I imagine a good portion of the population couldn't tell.

That saturation loss isn't caused by DV encoding. Well, at the very smallest scales it can be. Because of the 4:2:0 subsampling. But at larger scales it's not.

[sanlyn]

I don't know, png has always looked a bit grainy to me, a bit "hard". Less film-like ?? No loss from png copy to png copy, but an RGB original converted to png always looked "different" to me.

The h.264 examples in post #20 can be cpnfusing, because they display somewhat differently in the forum when seen reduced in size in normal view, but are quite different when clicked and viewed in their original size. The top image is darker, blacks are more crushed, brights are suppressed and midtones are more muddy. I don't think that's a compression problem, it's a conversion problem.

Cornucopia is correct, in that it's an example of converting but not an example of two ways of recording/capturing. There will be an easily visible difference between (a) capturing thru an all-in-one capture device with minimal image controls and tbc configuration, directly into a lossy compression format, compared to (b) capturing with the benefit of onboard hardware line-level tbc or via ltbc pass-thru device, then thru full-frame hardware tbc if needed, then a hardware proc amp such as SignVideo PA-100 at correct/adjusted luma levels, into a capture card/device designed to have a minimal effect on the images it transmits, and directly to lossless or uncompressed AVI.

I have no ready examples of method (a) because I never capture that way - I'm set up for (b) and have always captured VHS or "bad" video that way. It's a big pain and size problem for most people to post huffyuv or lagarith AVI captures.

[2Bdecided]

Was this originally shot to VHS or 8mm film? If film, find the original film, and start from that. A modern frame-by-frame pixel-sharp transfer straight from film will be 1000x better than some lousy old dub with a VHS camcorder.

If it was shot to VHS originally, you could re-transfer it with correct levels and a good TBC - either one built into the deck, or the kind built into a few DVD recorders or camcorders with analogue inputs.

DV vs lossless is a fantastically small difference compared to either of these steps. If the destination is DVD, then DV is good enough unless you have more noise than picture (which DV can't cope with) - it's good enough even as the source for extensive restoration. Lossless is better still of course, but I defy anyone to spot the difference once you've encoded the result for DVD.


If this was 8mm film transferred to VHS, and you've lost the original film, then after a TBC transfer, you should look at recovering the original film frames from the video using IVTC, or some of the blend recovery scripts in AVIsynth (e.g. restore24, with appropriate settings for the actual frame rate of the source), probably before doing anything else.

Cheers,
David.

[jagabo]

I don't know, png has always looked a bit grainy to me, a bit "hard". Less film-like ?? No loss from png copy to png copy, but an RGB original converted to png always looked "different" to me.

You can easily test this yourself. Take a 24 bit RGB BMP file, save it as PNG, open the PNG, save it as BMP. Compare the two BMP files. You'll find they're identical, bit-for-bit. I can't believe you're questioning this.

[sanlyn]

There's really not that much difference, jagabo, no dispute there. But I'll give it a run. In this specific case I see something happening with color space (DV->RGB), not compression or capture.

[jagabo]

I thought the firewire port on my old DV camcorder was dead. But I just tried it again on a different computer and I was able to capture from it (must have been a problem with the old computer). I'll see if I can capture some VHS tapes with both the DV camcorder (it has a composite input) and an ATI 650 USB2 capture device. I don't have an S-VHS deck with a line TBC, only old hand-me-down VHS decks. I do have a Panasonic ES15 DVD recorder with a line TBC so maybe I'll use that as a passthrough. I don't know what this will really tell us because DV camcorder has no proc amp controls (I don't remember how well it captures analog sources) and the ATI card is a very different device (it does support raw YUV 4:2:2 capture with lossless compression).

Oh, and all my equipment is NTSC, not PAL.

[sanlyn]

Good idea. If I had time (got 2 PC repair clients waiting for me) I'd cap straight to a DVD recorder, then same to AVI on the PC with the gear hooked up. The point is, capping directly to compressed video with no tbc, etc., just makes it tougher to clean and repair than capping to lossless AVI (compression artifacts of the former being the worst pain the neck).