Is an HD Super 8 transfer overkill?

[mattias]

I'm saying that upscaling SD to HD can work if the display isn't too large

i'm saying that *if* your source material isn't really hd res it shouldn't matter, it should look almost exactly the same scaled down and back up. if you see a difference it's because the source was actually sharper, or that your scaler was bad, or that you're fooled by the grain (try adding some back in digitally and see what happens?).

[Uppsala BildTeknik]

I was commenting on the fact that kent apparently hasn't, and more so that he doesn't even understand why it's necessary.

Funny thing, I am the only company, and as far as I know it the only person whatsoever, that has produced a testfilm to compare your typical SD and HD transfers and posted the results online.

I did it years ago, and the testfilm is online on my webpage. It has been there all along.

Where is your test? Or are you just talking about all this from a theoretical point of view?

This goes to everyone in this thread (except Roger, who has made tests). Where are your results/tests? Or are you all just talking from a theoretical point of view? Get your thumb out of your butt and make some real life comparisons. Unless you do, it is just all talk. :roll:

[Uppsala BildTeknik]

So Mattias, you will get SD transfers for your super8 stuff in the future. Correct?

there are other issues...720x576 pixel resolution and i get to chose...maybe 700x500 and records........

I´m not asking about more numbers, I am asking about if you are going to get your films transferred to HD or so SD?

Forget about what machine you will use. Do you think you will you get a SD or HD transfer?

If the answer is that you will probably get a HD transfer, then I can´t understand why you are arguing that HD is overkill? :?

[Uppsala BildTeknik]

... should look almost exactly the same scaled down and back up.

Oh, but if you are comparing whatever you want with a SD transfer there is no need to downscale the SD transfer. You just need to upscale it.

This is how you should compare a SD transfer to a HD transfer. By comparing... a SD transfer to a HD transfer.

[MovieStuff]

I'm saying that upscaling SD to HD can work if the display isn't too large

i'm saying that *if* your source material isn't really hd res it shouldn't matter, it should look almost exactly the same scaled down and back up. if you see a difference it's because the source was actually sharper, or that your scaler was bad, or that you're fooled by the grain (try adding some back in digitally and see what happens?).

"If, if, if, if, if..."

The way I see it, IF you transfer to HD to begin with, then you don't have to second guess how it will look later on. This holds true whether you are copying films off a screen with an HD camera or using the very best HD scanner known to man. But transferring to SD and then upscaling to HD invites too many unknown variables. So, in relation to the original question in this thread, I would say that transferring S8 to HD is not overkill and avoids a variety of unpredictable problems. Which problems does it avoid? We don't know, which is why they are called unpredictable. ;)

Roger

[Uppsala BildTeknik]

The way I see it, IF you transfer to HD to begin with...

Yeah, I can´t understand why everyone is using a HD transfer as a base that they downscale in order to prove that HD is overkill in resolution. Hell, if it was transferred to HD, why downscale it?

If people want to compare something in/from/to a SD transfer, they should use a SD transfer as a base for the comparison. IMHO. Otherwise, the comparison is just incorrect.

[MovieStuff]

The way I see it, IF you transfer to HD to begin with...

Yeah, I can´t understand why everyone is using a HD transfer as a base that they downscale in order to prove that HD is overkill in resolution. Hell, if it was transferred to HD, why downscale it?

To have a common point of reference. If the HD is downscaled to SD and then the SD version is upscaled back to HD and still looks the same, then that suggests that scanning in HD to begin with would be overkill. But the problem is that SD down-converted from HD always looks better than an original SD scan, at least to me, so you're upscaling from an artificially produced SD image and not from an original SD scan. The second problem is that SD upscaled to HD will NOT look as good as an original HD transfer if the HD monitor is large enough to justify HD to begin with. Regardless of what the numbers say it should look like, it just doesn't, no matter how the upscaling is achieved.

If people want to compare something in/from/to a SD transfer, they should use a SD transfer as a base for the comparison.

Agreed.

Roger

[carllooper]

Where is your test? Or are you just talking about all this from a theoretical point of view?

This goes to everyone in this thread (except Roger, who has made tests). Where are your results/tests? Or are you all just talking from a theoretical point of view? Get your thumb out of your butt and make some real life comparisons. Unless you do, it is just all talk. :roll:

I've done a real world test as well.

For comparison here is a Super 8 frame (TriX Canon 1014 lens), scanned at 3K.

Here we see it re-sampled at two different resolutions:

a. 2276 x 1688 (as in a scan for 200 lp/mm)
http://members.iinet.net.au/~carllooper ... ible35.jpg

b. 1138 x 844 (as in scan for 100 lp/mm)
http://members.iinet.net.au/~carllooper ... rior35.jpg

These demonstrate that scanning at 2K is better than scanning at 1K. But it does not demonstrate where the "overkill" point occurs.

However I can say I found the difference between 3K and 2K to be considerably less different than the difference between 2K and 1K.

There is an "overkill" point but I can't say where it is. All I know so far is that it must be beyond 3K.

If people want to compare something in/from/to a SD transfer, they should use a SD transfer as a base for the comparison. IMHO. Otherwise, the comparison is just incorrect.

Why is the comparison incorrect?

Carl

[Nicholas Kovats]

Carl,
Your prior hyperlinks did not reproduce correctly in your post.

Cheers

NK

[carllooper]

Hmmm. Copy and paste these:

members.iinet.net.au/~carllooper/super8/BestPossible35.jpg

members.iinet.net.au/~carllooper/super8/Superior35.jpg

[Uppsala BildTeknik]

Here we see it re-sampled at two different resolutions:

No, both images in your links seems to be 1920x1080 in resolution, and one seems to be quite soft if you look at the grain. Not sure if there is any difference in image detail. And again, you are using a higher resolution and scaling it back and forth... This will give you incorrect results, I´m not even interested to look at the images more closely.

I´m not sure what this test would show. It will not show what two different resolution transfers would give you. It will show what scaling back and forth would/could do to a image.

If people want to compare something in/from/to a SD transfer, they should use a SD transfer as a base for the comparison. IMHO. Otherwise, the comparison is just incorrect.

Why is the comparison incorrect?

Because it is. If you downscale a HD transfer to SD resolution, it will not show what you would get from a SD transfer.

If you want to see what you would get from a SD transfer, get a SD transfer. It will not look the same when compared to a downscaled HD transfer.

[MovieStuff]

If you downscale a HD transfer to SD resolution, it will not show what you would get from a SD transfer.

If you want to see what you would get from a SD transfer, get a SD transfer. It will not look the same when compared to a downscaled HD transfer.

Yes. I agree.

Roger

[carllooper]

Here we see it re-sampled at two different resolutions:

No, both images in your links seems to be 1920x1080 in resolution, and one seems to be quite soft if you look at the grain. Not sure if there is any difference in image detail. And again, you are using a higher resolution and scaling it back and forth... This will give you incorrect results, I´m not even interested to look at the images more closely.

The 1920 x 1080 is a crop - not a resampling - so that youcan fit the entire image on your computer monitor.

I´m not sure what this test would show. It will not show what two different resolution transfers would give you. It will show what scaling back and forth would/could do to a image.

It shows that a 2K resample resolves more detail than 1K resample. When you transfer a film to digital you are doing the same thing - resampling a higher definition image (the original film).

Why is the comparison incorrect?

Because it is. If you downscale a HD transfer to SD resolution, it will not show what you would get from a SD transfer.

Why not?

If you want to see what you would get from a SD transfer, get a SD transfer. It will not look the same when compared to a downscaled HD transfer.

Why?

And here is the difference between the two resamplings:

http://members.iinet.net.au/~carllooper ... erence.jpg

If the 2K sampling was "overkill" it would just show noise - but as can be clearly seen, the teapot is still visible. The point at which "overkill" occurs is when you can no longer make out the teapot (the signal), ie. where you can only see noise.

Carl

[Uppsala BildTeknik]

Why?

Why is the sky blue? Or is it blue?

Who cares. The sky is not blue and a SD transfer will not look the same if compared to a HD transfer that you downscale.

If you don´t believe me, order a SD transfer from a company that can deliver SD transfers and upscale it yourself. Compare it to a HD transfer. They will not look the same.

Does it matter why? If it does, you can go into all kinds of details about this, I guess.

To me, it doesen´t matter. I know it will not look the same, and that is enough. I feel no need to know all the nit-picking details about why it is so, and I don´t have time to nit-pick all the details about all the parts of every step. I know the details, but I don´t feel the need to know the details about the details about the details. :mrgreen:

[MovieStuff]

Because it is. If you downscale a HD transfer to SD resolution, it will not show what you would get from a SD transfer.

Why not?

Because down scaling from HD usually produces better results than capturing in SD to begin with.

If you want to see what you would get from a SD transfer, get a SD transfer. It will not look the same when compared to a downscaled HD transfer.

Why?

Would you really think a Rank transfer to SD will really look as good as an HD Spirit transfer down scaled to SD?

One could take the position that the SD transfer should be done on the same caliber machine as the HD unit but the fact that you have to make such a distinction only emphasizes the variables inherent in SD > HD scaling. If you transfer in HD to start with, you don't have to make such considerations. In that context, transferring S8 to HD isn't overkill at all; it's just common sense that avoids unexpected issues inherent in SD transfers.


Roger