What do these 3 exported frames say to you?

[guyburns]

Comments on the results of some test captures, please. They're of Jessie Vonk being interviewed for a news item in St Louis, around 1996. Equipment used:

• JVC HR-S6960
• Panasonic DMR ES15
• AJA Io LA capture unit
• Late 2013 iMac running OSX 10.9.5 and AJA VTR Xchange capture software

Attached is a zip file, and inside is a tiff image with 3 aligned layers. You'll need to open the file in PS and turn the layers on and off to compare the captures. The layers were exported from Premiere, unrendered, so they're are MPEG 2 quality, not full quality.

The three layers are of a raw capture; a raw capture using the DMR ES15 for passthrough; and one of those has been edited to make it look like the other.

I'm not asking for a vote, or which is the best result. I'm asking for comments that may enlighten me for when I finally put together my analysis of various capture methods.

[Gary34]

You might get a no response with that. Well, besides me now. I don’t think that they are going to go to photoshop and compare those. If they think your method of coming to your conclusion is flawed then they might tell you why but you will probably think those are theoretical reasons. They might just not comment.

They may not get into it though because I suspect some of the details with using Adobe as a way of comparing raw files might be a little bit of a worm hole in itself.

I suspect that that isn’t the best way of comparing different methods. I think your color space is going from YUY2 to Adobe RGB, then you well go to sRGB I believe and I don’t think Adobe handles colorspace conversions as well as hybrid or Avisynth. Also you are using a picture to compare video and thinking that what you can do in software if a replacement for hardware changes.

Also with compression it behaves differently in editing.

The losslessley compressed codecs behave way different and the software they use to handle those codecs is a lot different. A visual doesn’t show that.

[Aya_Rei]

As Gary34 said, Still images aren't enough here, with short, small raw video samples being needed. Looking through the two images myself the only things that come to mind is "Wow, the ES15 sure gave the image a redder tint and added some color compression, especially to the background" if layer 2 is using the ES15 and layer 1 isn't using it.

I might not be the best person to point out differences between using the ES15 and not using it unless it's super obvious. I might be wrong, but since the ES15 adds it's own issues with the positive being able to correct jitter and tearing, it'd be best to not using the ES15 unless it is fully required. Adding more components to a workflow creates more variables over the end result.

[traal]

I agree, lossless clips would be better. Then I could put them through my software workflow and see which one comes out best. To me, that's the real test.

[Gary34]

A little off topic but I think I messed up when I was talking about mpeg2 colorspace.
mpeg2 and mpeg4(dvix,xvid) encoders do in general encode in YV12 means 4:2:0 - caution in case of fieldbased capturing!
- mjpg encodings will be most done in YUY2 - 4:2:2
- HuffYUV does encode in YUY2 - 4:2:2

I compared some videos today in Pmod with the stack horizontal function just to get an idea of what my tapes look like with and without my TBC turned on in my Panasonic 1980p ag and these tapes look really good with my TBC turned off. I well need it on for some of my other tapes though. My TBC 3000 is far more transparent. That could be because of caps or some other combination of something who knows. It’s crazy how well Pmod handles all that.

My point is if you aren’t using there software you are shooting yourself in the foot worse than you think.

If it was me I wouldn’t wanna start this project not using there software and codecs. They have this all figured out really well. I have the same computer hardware I had before I saw this site and now everything goes super smooth.

[latreche34]

3 layers? Haven't heard of that yet, I don't own any software that either requires me to pay an arm and leg or a monthly subscription.

[lordsmurf]

Quote:

Originally Posted by Gary34

using Adobe as a way of comparing raw files might be a little bit of a worm hole in itself.

Correct, colorspace.

[Aya_Rei]

If anybody would need the images in a forum post, then here they are

No ES15 -
Jessie_NoES15.jpg

ES15 -

You must be logged in to view this content; either login or register for the forum. The attached screen shots, before/after images, photos and graphics are created/posted for the benefit of site members. And you are invited to join our digital media community.

Combination -
Jessie_WithES15.jpg

[lordsmurf]

These aren't the same frame.

The top frame looks like DV (color loss), maybe ES15 (smooth). But too much chroma noise.

ES15 has known luma over/under exposure. So second/middle could be. Exposure alters saturation.

Bottom image seems to have chroma leak from adjacent frame.

Hard to say too much, observations could be off. Details needed, clips needed.

I hurt too much to type more, all you get.

[Aya_Rei]

To me anyway, it was hard to determine which image layer was the ES15 and which image wasn't, the only one I was certain about was the third image layer being the combination image. So I'm sorry if I mixed up what image is meant to be the ES15 and what image wasn't it. I just went by with the order the OP listed, thinking that the first image layer is the one without the ES15 in the chain, and the second image layer being the one with the ES15 in the chain

Though if lordsmurf's reply was more directed towards guyburns and not me (which it seems to be from his comment on requesting video samples) then I apologize.

[lordsmurf]

Quote:

Originally Posted by Aya_Rei

if lordsmurf's reply was directed towards guyburns

^ This.

.... but based on your posted images, I can't handle layered PSD right now.

[guyburns]

Hi everyone, and thanks for the comments.

Static Comparisons
I have chosen to do static comparisions as a first step, because it is impossible to accurately compare moving images if the captures are close in quality. The only way to compare captures, when trying to determine the most accrate method of capturing, is to use multiple layers in Photoshop, with the various captures aligned, so that when you change layers, it becomes very obvious what the changes are between layers. It is simply not possible to accurately compare static captures if you open those captures in a different physical position. Comparisons have to be done statically, have to be aligned, and when a certain capture method has been decided upon, only then is it a sensible idea to compare videos.

In future, I can see that I'll have to preface any more comparisons with the comments:

1. There is little point in contributing to this thread unless you have Photoshop and know how to turn layers on and off.
2. If you do have PS, there is no point in extracting the layers and uploading them here, because that will introduce all sorts of changes. Layers have to be compared within the same file so that there are no alterations to those images caused by copying and pasting.

Comment 1
The kind of comments I am seeking are like those below, which I obtained from two friends involved in the local camera club for the last 40 years, one of whom did his thesis on something to do with TVs way back in the 70s.

• Your captures. Looking at them filling my 4K monitor they look very good considering its from a VHS tape. In terms of resolution, noise, and artifacts I can see little difference between them.
  
• Colour… The actual colour hue, determined by the carrier phase is good. PAL is optimised to get good skin tones. NTSC was notorious for colour cast on the skin.
  
• However, I can see a difference in saturation between the images. Colour gain or saturation can be an issue for PAL TV as its amplitude modulated on the colour subcarrier. Increasing the gain increases the noise. Higher colour gain ( saturation) also gives skin tones a sunburnt look.
  
• My Choice: Layer 1 is the least saturated and most pleasing to me, followed by Layer 3 with Layer 2 being the least appealing. As you switch between the layers in the order of my choice you may just see the colour noise increases on the cheeks.

Exactly! He detected that Layer 1 was best, being raw from the AJA; that Layer 3 (via the ES15, but corrected for colour) was close; and that Layer 2 was the least desirable (unedited from the ES15).

What he didn't detect (I only saw it after uploading), was that the AJA has a slight problem with line sync: the left edge is ever so slightly bent to the right. Layers 2 and 3 correct that bending. Try and see that difference outside of a layered Photoshop image!

Comment 2
Guy, looking at them closely and enlarged to 250%, it is a close call between Layer 1 and 2. It appears Layer 2 may be a little sharper than Layer 1, however I feel Layer 1 has better colour and exposure. Layer 3 has the poorest colour and sharpness. This is splitting hairs, but enlarged to 250% this is how it appears to me. This is an interesting exercise in image post processing. I must call by soon so we may talk face to face.

So, both friends agree that Layer 1 is the best, but disagree on second place. That's where personal preference comes in, but I'm not interested in that at this stage. I'm asking for comments on my captures that detail the technical differences, why they might have come about, and how they can be corrected.

[Feedbucket]

What was the intention behind doctoring the 3rd layer to look like the first?

Also a little surprised that 2 and 3 are ES15 output as it seems to be doing the opposite of what it's expected to. The vertical edges don't appear to be aligned and the lady's jawline is distorted.

[guyburns]

Quote:

Originally Posted by Feedbucket

What was the intention behind doctoring the 3rd layer to look like the first.

The intention is to see if I can make an inferior capture look the same as a better capture. If that's possible, then as far as I'm concerned, they are identical. i.e. if I can make an Elgato look like a AJA, why spend all that money on fancy capturing equipment?

I did the same sort of thing when I was testing slide scanning. Even though the Nikon Coolscan (a dedicated slide scanner at 4000dpi) gives sharper scans than the Epson V700 (the best-quality flatbed for scanning slides), the Coolscan gives sharper results because there is no glass between the slide and the scanning element. However, Epson scans can be made to look virtually identical to Nikon scans by sharpening in Photoshop. But you have to know how to do that – scan at 6400 dpi, reduce to 3200 dpi, and apply a certain amount of sharpening. Only by detailed examination inside PS was I able to prove that result.

For my slides destined for AVs, I only use the Nikon. But if a friend wants me to scan some of their slides, I use the Epson, mainly because I can scan 12 at once, and then apply the editing via an action.

To reiterate, I'm testing VCR captures to achieve the best result when finally projected on screen. The AJA IoLA is very good for sharpness and colour, but it requires time base correction. The ES15 can act as a very capable TBC, but the reds bloom. Layer 3 shows the result of applying colour corrections in Premiere to tone down the reds, to bring them into line with the raw AJA scan. Layer 3 was a quick edit in Premiere, but now I jump inside After Effects from Premiere to make use of AE's fancier colour tools, which Premiere lacks. I'm hoping I can make the ES15 captures look better than the AJA alone.

That comparison will be the topic of an upcoming thread.

P.S. The 3-layers I uploaded, just one frame chosen at random, are not necessarily representative of the entire video. I've noticed that if I move one frame forward or back, Layer 3 can appear sharper than Layer 1. That tells me the supposed extra sharpess I think I'm seeing is not extra sharpness at all, but an artifact of more "spots" appearing on the frame for whatever reason, similar to sharpening a slide where the grain becomes more visible. You think it's sharper, but is it? That's another aspect I've yet to look into. And that's the point where I will have to start watching videos.

[lordsmurf]

Quote:

Originally Posted by guyburns

. PAL is optimised to get good skin tones. NTSC was notorious for colour cast on the skin.

This is decades-old myth/BS. PAL looks as craptastic as NTSC on skin tones. The digital age really betrayed that dogma as the complete nonsense it always was. PAL color was just wrong differently, not more accurate.

Quote:

pleasing to me

So this is the type of comments you're seeking? Forget science, just "me likes it".

Quote:

What he didn't detect (I only saw it after uploading), was that the AJA has a slight problem with line sync: the left edge is ever so slightly bent to the right. Layers 2 and 3 correct that bending. Try and see that difference outside of a layered Photoshop image!

Aja cards are not TBCs, and do not (meaningfully) contain TBCs. No mystery there.

Quote:

That's where personal preference comes in, but I'm not interested in that

But .... huh? All of your buddy comments were just random opinions.

Quote:

I'm asking for comments on my captures that detail the technical differences, why they might have come about, and how they can be corrected.

And that requires exact knowledge of a full workflow: VCRs, TBCs, capture cards, software, and anything else. This stuff should not, and cannot not, be a guessing game. Otherwise it's just "me likes it" and "that one looks gooder" comments.

[Gary34]

Quote:

I compared some videos today in Pmod with the stack horizontal function just to get an idea of what my tapes look like with and without my TBC turned on in my Panasonic 1980p ag and these tapes look really good with my TBC turned off.

My opinion changed on that today. I took a second look and saw the line (field) TBC in my Panasonic reducing color bleeding quite a bit. I am going to capture with line and field TBC on all my tapes.