My end goal is converting 8mm (Video8) analog tapes to digital files of good quality.
This is the recording, playback, and capture information:
Original recording analog on Olympus VX-801 (Video8)
Playback on Sony DCR-TRV350 (can play analog or digital)
Capture Card (used on 2 of the 3) - Magewell Pro Capture HDMI
I was not sure how to do this... tried to create animated GIF in GIMP, but colors were "blocky" (don't know the technical term... but not gradual) and gave up. Ended up with this .avi out of VirtualDub (attached). I didn't try to figure out how to lower the frame rate, so you will have to step through. I was trying to figure out a good way to compare the frames... If anyone has efficient ways to compare formats, I would love to hear about those...
The 3 formats (3 frames in the avi):
0. Component Composite video out of the old Olympus (YUY2)
1. DV out of the Sony (via firewire)
2. S-Video out of the Sony (YUY2)
Obviously, the detail increases dramatically from the Component Composite capture to the other two formats. However, the noise (is that the correct term) is much worse in both the DV and S-Video captures. However, the Olympus is pretty old, and magnification of the tape heads seem to show roughness (maybe corrosion?). For that reason, I don't know if the Component Composite capture is a fair representation.
I am trying to define my workflow before diving into this project. I guess the main question is: Can the S-Video be cleaned up to look as good as the Component Composite Video capture with respect to "noise."
Any advice is appreciated.
Edit: Sorry (dyslexia? - Composite, NOT Component)
+ Reply to Thread
Results 1 to 30 of 33
Last edited by GrouseHiker; 10th May 2020 at 11:00. Reason: Renumbered frames starting at 0; "Composite," not "component" :-(
You can't get more than what's already recorded on tape which is a Y/C signal, Such signal is output via S-Video. Y/Pb/Pr, Composite and iLink are just further processing of the original Y/C signal. What some people call noise is just more details that other methods of processing will wash out the colors and smooth out the details resulting in a soft looking video.
Component Composite-video capture was it is substandard... until I noticed the noise difference. The Component Composite video somehow looks like a higher pixel density in the field of the skin - blended noise rather than discrete noise?
With the proper tools and proper application, will I be able to produce (from the s-video capture) the clean look of the kid's skin that shows in the Component Composite-video capture? I understand there's more to video quality than the "soft" skin but it does look realistic in the component-video frame.
Last edited by GrouseHiker; 10th May 2020 at 10:03. Reason: "Composite" NOT "component"
What kind of component output is it? Never seen component on a consumer camcorder especially that old, Can you post a picture or a manual page?
The best that Olympus cam can put out is s-video (or 1/2 component, if you will) analog out via the multi-pin mini-din jack.
The best the Sony can put out is s-video (if you go analog) or dv (if you go digital via firewire).
Neither of those will natively connect to your Magewell card, which is HDMI on the input.
To go dv you would need a firewire port, or card with a fw port, if you don't already have a port on the mobo.
This bypasses the magewell entirely. And you are (probably unnecessarily) sticking the 1st generation of capture with dv compression. Not the most optimum start.
Going via s-video you need a capture card that has s-video in. Also not really using the magewell. Or if you found some s-video --> hdmi adapter, you could use it, but you would be at the mercy of the quality of the adapter.
Kind of a rube goldberg approach.
The hdmi is the only one of these formats that is fully component, but none of your sources support it.
The olympus cam, if it was the original cam that made the tapes, would probably be the best playback device.
However, if it has suffered some damage, or even some major wear and tear, all bets are off.
Btw, it doesn't hurt to do mutiple passes and do a median-type noise reduction process on them (though you must of course EXACTLY match the footage if you do try that), and you could even include playback via both cams.
For optimum quality, you want the original capture(s) to be as linear, straigh-laced & unadorned as possible, using proper levels. Save as losslessly-compressed or similar, Then you can apply what nr or enhancement processes you feel are warranted.
I think that Magewell HDMI card has analog inputs as well but I'm pretty sure the analog section is an after though, I still want to know what kind of component output that Canon camcorder has, Not something that I'm aware of in consumer camcorders, I started to think he meant composite.
Last edited by dellsam34; 10th May 2020 at 01:48.
Yea, I think he means composite.
The Magewell is pretty solid for s-video captures. While the manufacturer support says there's no TBC enabled, the captures say otherwise. Lines are stabilized and it can handle discontinuities in video without dropped frames (rough slices, fast forward etc). I've not personally tested, but others indicate it does not seem to have its 3D comb filter enabled or tuned properly for composite sources. You just want to make sure your proc amp controls are set properly, pedestal matching your source, and disable deinterlacing.
Like previously posted, best setup (assuming playback devices are in similar state of repair) is s-video. I'm a little paranoid and without confirmation from a technical manual, I went for a Hi8 camera instead of a Digital8 camera just in case any video signal was routed thru its digital circuits and resulted in a A->D->A->D conversion. Probably not the case, but one more playback device was cheap. You can gauge the level of paranoia in that I avoid/disable all TBCs/DNR as many of that era were 8 bit or less, and I want my analog signal, stealing from Cornucopia, "straigh-laced & unadorned as possible" before it hits my ADV7842. After that you can post process anywhere from vaporwave to plasticy smooth.
The Olympus adapter (VF-BA81) has composite out and a coax port out (which has been destroyed). The camera connects through this port (photo).
[Attachment 53217 - Click to enlarge]
The Sony has s-video and DV out.
The Magewell card has analog inputs via a DB9 adapter. IT accepts Component, but I don't have any Component player outputs:
Analog audio (L+R)
WinDV or Windows Media Encoder.
jagabo babygdav (https://forum.videohelp.com/threads/394319-DCR-TRV350-to-PC/page2#post2580589) said not when I asked the question.
I do plan to do testing with camera TBC and/or DNR off.
Last edited by GrouseHiker; 12th May 2020 at 10:45. Reason: reference regarding camera processing of analog was babygdav, not jagabo
I'm definitely planning to save the Archive and "Mastered" copy (if needed) as interlaced lossless. I was planning to not touch the Proc Amp setting.
VirtualDub is supposed to be able to adjust to dropped frames.
I think the "alligator" skin problem you're having isn't on the tape. And it's not caused by composite video. It's some kind of electronic noise in your capture setup. A ground loop problem maybe.
[Attachment 53262 - Click to enlarge]
I was thinking maybe the digial camera used for playback is creating this, since the composite capture off the original camera doesn't have it (attached).
To simulate, I tried shrinking the component-video capture horizontally and then re-expanding it:
AssumeTFF() SeparateFields() widthstart=width BilinearResize(width/5,height) BilinearResize(widthstart,height) Weave()
Still searching for a different playback device. I'll buy one off eBay if I have to, but I'm trying to stay away from a digital camcorder - maybe a Sony CCD-TRVxx.
#BASED ON: [QUOTE=jagabo;2582052] # https://forum.videohelp.com/threads/396961-Weird-Lines-in-8mm-Captures/page3 AviSource("1986_1011 Pool Sample3 YUY2.avi") #WEIRD LINES REMOVAL --------------------------------------------------------------------------------------------------------------------- AssumeTFF #YUY2 #AssumeBFF #DV SeparateFields() AssumeFrameBased() AssumeTFF SeparateFields() modEven = SelectEvery(4, 0) modOdd = SelectEvery(4, 3) okOdd = SelectEvery(4, 1) okEven = SelectEvery(4, 2) Emask = modEven.ColorYUV(off_y=-120).ColorYUV(gain_y=512) modEven = Overlay(modEven, modEven.ColorYUV(gain_y=-7),mask=Emask) Omask = modOdd.ColorYUV(off_y=-120).ColorYUV(gain_y=512) modOdd = Overlay(modOdd, modOdd.ColorYUV(gain_y=-6),mask=Omask) evenlines = Interleave(modEven, okEven) oddlines = Interleave(okOdd, modOdd) Interleave(evenlines, oddlines) Weave() #AssumeFieldBased() #Weave() #RESTORE SECTION ------------------------------------------------------------------------------------------------------------------------------ #jagabo post#11 #https://forum.videohelp.com/threads/393688-Update-18-07-Restoration-Tips ConvertToYV12(interlaced=true) #AssumeTFF() #SeparateFields() vInverse() # reduce combing from time base errors AssumeFieldBased() Weave() Levels(16,0.950,235,10,250,dither=true,coring=false) Crop(14,0,-8,-0) # maybe the bottom too, to get rid of the head switching noise FineDehalo(rx=3.0, ry=0.1, darkstr = 1.0, brightstr = 1.0, showmask = 0) # reduce over sharpening halos QTGMC(InputType=0, sharpness=1.0, EZDenoise=7.0, DenoiseMC=true, SourceMatch=3, Lossless=2) # clean up edges, denoising (, preset="Fast") #sRestore(frate=59.94) #29.97 119.88 MergeChroma(last, aWarpSharp(depth=15)) # sharpen chroma #MCTemporalDenoise(settings="low",enhance=true) #gradfun2db.dll only 32 bit ColorYUV(gain_y=0, off_y=0, gamma_y=0, cont_y=0, gain_u=0, off_u=0, gamma_u=0, cont_u=0, gain_v=0, off_v=0, gamma_v=0, cont_v=230, levels="", opt="", showyuv=false, analyze=false, autowhite=false, autogain=false) AssumeFPS("ntsc_double", sync_audio=true) #Histogram("classic")
I cranked up the "EZDenoise" parameter in QTGMC, and the alligator skin was reduced, but it started doing damage elsewhere.
I could not get the playback speed adjusted back to where it should be.
[Attachment 53265 - Click to enlarge]
Am I on the right track????
I wouldn't recommend using QTGMC's EZDenoise over 2.0 or so. It will lose detail create smeary artifacts where there's motion. This is where you would be much better off getting caps without that herringbone noise. Try things like running the camcorder off the battery -- so it's not plugged into the power mains. Or plug the camcorder and the computer into the same power strip. Try moving the Magewell card to a different slot. As far away from any other cards you may have installed. There are a lot of noise sources inside a PC. None of the Magewell cards I've seen have any shielding. And, of course, try another playback device.
ColorYUV(cont_v=230) is too saturated, I think. The video needs some white balance too. Hard to do well in AviSynth.
I tried playback on battery - same "herringbone" (vertical crawing worms). The composite capture off the old camera went through the same Magewell card, and that capture didn't have the worms. I'm pretty sure it's the camera at this point... looking for a different player...
Regarding "white balance" (if I understand the goal of proper white balance), I was torn between opposing goals. I was thinking the video should look like it was shot in the bright Florida sun, but that washed out the detail in the pool coping for example. I had gone for washed out first and then changed it.
Last edited by GrouseHiker; 12th May 2020 at 21:34.
This might be going a little too far but... partly in AviSynth, partly in VirtualDub:
[Attachment 53284 - Click to enlarge]
Note, I didn't remove the horizontal lines or make any attempts at noise reduction. Just white balance and levels adjustments. In a shot like this I assume it's more important to see the kids' faces, not the background. Also, I guessed that the concrete around the pool was white.
Last edited by jagabo; 13th May 2020 at 08:47.
That looks good. I tried pushing the output level way high (Levels(16,0.950,235,16,300,dither=true,coring=fal se)), but that didn't seem to work.
How did you achieve that?
I didn't save the VirtualDub settings (I used the Gradiation Curves filter to bring out shadow detail) but I roughly replicated it in AviSynth:
ConvertToRGB(matrix="pc.601") RGBAdjust(r=245.0/208.0, b=238.0/250.0, rb=-5, bb=7) # white balance RGBAdjust(rg=1.5, gg=1.5, bg=1.5) # gamma to bring out shadow detail in the faces ConvertToYV12(matrix="pc.601") ColorYUV(gain_y=-13, off_y=-10, cont_u=77, cont_v=77) #Tweak(cont=.95, bright=-10, sat=1.3) # alternate for above ColorYUV using Tweak instead
https://forum.videohelp.com/threads/287113-Gradation-Curves-VirtualDub-filter-v1-45), I got to here (http://members.chello.at/nagiller/vdub/tutorial/tutorial.html), which is pretty amazing.
[Attachment 53346 - Click to enlarge]
If I'm reading this correctly, I think the "alligator skin" / "herringbone" / "vertical worms" noise is "Mosaic Noise." It looks like mosaic noise is a digital camcorder (playing Video8) issue, but I'm not positive. I didn't see any mention of mosaic noise in a pre-digital, Hi8 Operating Manual.
The experimental eBay camera came in (CCD-TRV318), which happens to be the same-year little brother (older brother?) to the DCR-TRV350 I have. Unfortunately the Hi8 player version produces the same noise patterns as the digital camera as seen in the 720x480 capture:
[Attachment 53585 - Click to enlarge]
I tried capturing at 1440 x 480 to see if higher resolution would change anything... but no obvious change other than the noise being twice as wide and possibly better resolution of the noise. (I have no idea why the color is different...):
[Attachment 53587 - Click to enlarge]
The vertical-ish artifacts tend to be linked "worms" in the original, interlaced frame, but sometimes they look like hieroglyphics... there is no repeating pattern than can be calculated.
The composite capture off the original Olympus camera (original post) looks much better with regard to this noise, but the images are pretty fuzzy. I'm thinking the best option is to try to work with the s-video capture.
In the images, you can see the noise tends to be horizontal groups of pixels, one pixel high. Is there a denoiser that only compares vertically and eliminates aberrant pixels that don't match the pixels above and below (if the above and below pixels match one another within reason)?
A clip from the Hi8 player is attached.
Last edited by GrouseHiker; 29th May 2020 at 10:28.
I don't know why you're getting that noise. If it was composite video I'd assume it has something to do with chroma carrier, dot crawl artifacts.
Have you tried shooting a new short video and capturing it? Does it show the same noise?
Yes, same noise. Can you try a bright outdoor shot? To rule out gain noise. Though I guess it doesn't matter. Your existing tapes are what matters.
I was also thinking your older tapes might be second generation -- recorded via composite from another tape. But the fact that a new recording has the same noise eliminates that possibility.
I was looking at a closeup of the water, and it looks like there is dark and light noise that seems to have the same shape. I was searching for a denoiser that is spatial - vertical only option - but I couldn't find anything that seems obvious. If course my knowledge of this subject is VERY limited.
I think it's imaging sensor related, Some first generation camcorders used a very limited pixel count and they interpolated the resolution coming out of the censor processor.
[Attachment 53609 - Click to enlarge]
I have searched and searched for an Olympus Hi8 as a possible playback option, but I can't find any such camcorder. Olympus must have had limited or no production of Hi8.
I was also wondering if maybe a pre-digital era Sony (1997-ish) might process the signal from tape differently.
Last edited by GrouseHiker; 31st May 2020 at 11:16. Reason: attached png of composite capture
The composite capture just blurred it all away while separating the luma and chroma carrier. You can get similar blurring in software with something like:
Blur(1.5, 0.0) Blur(1.5, 0.0)
Speaking of TBC, I noticed the Composite capture off the old Olympus (no scan-line TBC) has what appears to be scan-line drops (white streaks). Is this the result of no scan-line TBC?
Also, does "pedestal matching your source" mean line count = 480?
This is the "Input" tab under "Capture Filter" in VirtualDub64. I understand the 525 vertical, but not the 858 horizontal:
[Attachment 53610 - Click to enlarge]
Last edited by GrouseHiker; 31st May 2020 at 11:59.