18 fps

[tlatosmd]

You lose exactly as much information as by a lower shutter speed of 25fps as compared to a higher shutter speed of 50fps. That's what that 'blur' is, it's exclusively a temporal motion blur like a lower shutter speed, not a lower pixel resolution blur.

As for that slomo, it's one of the methods a co-student and I used for a dissertation on different ways of achieving a smooth slomo. Your interlacing pattern looks like AB, AB, AB... in the original footage (A being odd lines, B being even lines). You deinterlace your footage two times, resulting in a progessive file that has progressive A material, the other file containing all the progressive B material. Then you'll re-sort those frames so the progressive frames will *alternatingly* go A, B, A, B... (as opposed to original AB, AB, AB... fields displayed *at the same time*). What you'll get is a progressive video shot at 50 deinterlaced frames/second and played back at progressive 25fps, realllly smooth.

However, this slomo actually gives you half the image resolution as the information from the complementary field has been discarded, it hasn't been used. That's why the deinterlaced A file and the deinterlaced B file look pretty much squeezed together and you'll have to stretch them back to the correct ratio before combining them to the final A, B, A, B... file.

I'll try getting one of those slomo files online to show you. You'll see how good resolution looks like, and it's even at half the pixel resolution of an ordinarily deinterlaced file (because the slomo file alternatingly discarded A and B field information during deinterlacing while ordinarily deinterlaced footage uses *both* fields) because we cheated a bit during correcting ratio, we used a bilinear or bicubic resize (which still gives more blurred images in *PIXEL RESOLUTION* than ordinary deinterlacing using *BOTH* fields). That slomo only looks a bit jumpy like a bad registration, the picture alternatingly jumps up and down one line.

However, ordinary deinterlacing of an interlaced AB, AB, AB... file doesn't really result in A, B, A, B..., or AA, BB, AA..., or whatever, but rather C, C, C... with C being a *combination* of A *and* B.

[tlatosmd]

I'm sorry for keeping you waiting for so long Kent, I had to find a host for these vids.

First, let's have the original, interlaced PAL DV file: http://media.putfile.com/Salto-links-QDV (6.04MB).

Now, the slomo file created by using deinterlaced odd and even fields recombined, making up a smooth 1:2 slomo: http://media.putfile.com/Salto-links_50-HBI71 (4.35MB).

Let's have a little bonus, two S8 versions of this scene ;), made to compare digital ways of achieving slomo with an analogue one (shot with my Bauer cam at 40fps on Ektachrome VNF 7240): http://media.putfile.com/Salto-links-S8 (5MB), http://media.putfile.com/Salto-rechts-S8 (5.05MB).

Lastly, let's have a look at what a deinterlaced PAL DV file looks like as for resolution: http://media.putfile.com/DV_deinter (9.35MB; please keep in mind this file went through MicrosoftDV and then was converted to XVID to squeeze it below 10MB in order to being able to host it).

[tlatosmd]

Posted 4 times while the server was down! 8O

[tlatosmd]

Posted 4 times while the server was down! 8O

[tlatosmd]

Posted 4 times while the server was down! 8O

[Uppsala BildTeknik]

Hmmm, those two files seem to have different resolutions? At least when I click the "view as original width" I see two different sizes... If so, the comparison is not that good you know...

Also it would be better if they were both downloadable, so everyone could download them and then view them in thier NLE and export still frames or whatever. Now I could right-click on one of them and save it, but the other I couldn´t. :?

It was the interlaced "original" file I was able to download, and what do I see: well how is it possible that the girls left hand is visible three times on frame nr. 17? It should only be two times (one for each field).

Also what is this shot with?

If Super8, why are all frames interlaced ? How was it transferred?

If DV, why is there a pulldown (some frames are repeated)?

If you re-render and compress the files a lot they are no good reference for sharpness anymore... Nice slow motion though!

But for a sharpness comparison I guess still frames would do, if they are from the same frame, it would save the trouble for hosting the files.

[tlatosmd]

The first two files are both 720*576 (non-square). The two following after that are two S8 clipss, both having a resolution of 720*530 (square), as the the latter are not to compare with the first two, they're only a little bonus. We color-corrected the first two files based on the S8 files as we scheduled to find out about the quality of particular slomo methods by surveying unbiased co-students who didn't know what methods we'd used (S8 won hands down in the ratings ;) ).

The first file is interlaced as that's how it came from the tape, the second file is a bit more blurred than regular de-interlacing like I've told you:

You'll see how good resolution looks like, and it's even at half the pixel resolution of an ordinarily deinterlaced file (because the slomo file alternatingly discarded A and B field information during deinterlacing while ordinarily deinterlaced footage uses *both* fields) because we cheated a bit during correcting ratio, we used a bilinear or bicubic resize (which still gives more blurred images in *PIXEL RESOLUTION* than ordinary deinterlacing using *BOTH* fields).

That's why it looks like it contains a pulldown.

Here's a frame from the first, interlaced file as it came from the tape (shot with a DV cam), saved in a DV mov file:


And here's the same frame from the second file, the slomo, saved in the same format:


As for the sharpness of deinterlaced video in general, here I posted frames of the very same project my last clip is from, and it's all deinterlaced footage: viewtopic.php?p=94147&highlight=college#94147

Actually, deinterlacing usually is *giving back* the image its pixel resolution that it lost by interlacing (remember, interlacing was invented to divide the resolution in half in order to make the cathode ray faster), giving it back by using the increased temporal resolution.

[lastcoyote]

If you want to make 25 frames from a source that is 18fps you would either need to copy some frames that makes the image look jumpy. What most scanners do is to create new interpolated frames that consist of two different frames. This will make the film look slightly blurred when played back.

So what you mean if I make 29.97 frames from a source that is 18fps... It will even worst?

[VideoFred]

Does no one use Motion Perfect, here?
http://www.dynapel.com/products/mp_prob_sol.shtml

Fred.

[Uppsala BildTeknik]

Actually, deinterlacing usually is *giving back* the image its pixel resolution

Seriously, de-interlacing is increasing and improving the resolution, hsarpness and quality of the images? I will never ever believe that.

OK, so they are transferred "off the wall" with a miniDV camera. Too bad there is no "clean" images with no interlacing artefacts, all those interlace lines makes it pretty much impossible to compare sharpness, everything is full of lines.

You should take a "full frame" (no interlace fields visible) and de-interlace it and then compare it to the original interlaced image, then you have comparable images to compare sharpness with.

[tlatosmd]

Does no one use Motion Perfect, here?
http://www.dynapel.com/products/mp_prob_sol.shtml

Fred.

We did use that as well. It looked mostly psychedelic.

Seriously, de-interlacing is increasing and improving the resolution, hsarpness and quality of the images?

Sharpness and especially quality might be highly subjective terms (as interlacing lines might be regarded as a distortion, kind of blurredness, lack of visible detail), however deinterlacing is increasing in pixel resolution by decreasing temporal resolution, and I've given the explanation, an interlaced field was *designed* in the late 1920s to be half the information of a progressive field so the cathode ray could reach its starting point in time before said point went dark again.

OK, so they are transferred "off the wall" with a miniDV camera. Too bad there is no "clean" images with no interlacing artefacts, all those interlace lines makes it pretty much impossible to compare sharpness, everything is full of lines.

I wouldn't know if the above S8 examples were shot 'off the wall', they were transferred by Andec. But it doesn't matter, as I didn't post them to compete with the posted DV files, I posted them exclusively because they are S8 and this is what this forum is about, so their quality doesn't matter as they're not supposed to compete with the DV files here.

Or do you mean we shot the two DV files 'off the wall'?

You should take a "full frame" (no interlace fields visible) and de-interlace it and then compare it to the original interlaced image, then you have comparable images to compare sharpness with.

That's why I gave you that link to those screenshots of my other project, to give you an idea of how sharp deinterlaced footage can be. Do they look *sharp enough* for you? That's the decisive factor.

I might dig up some examples of the same frames interlaced and deinterlaced if you ask me to but from my experience they blur *EXCLUSIVELY* in parts that were moving within 1/25 seconds, which is the completely normal motion blur you get when setting your shutter to that speed. Deinterlacing is exclusively a perfect simulation of a 1/25 seconds shutter speed on footage shot at a shutter speed of 1/50 seconds.

I ask anyone here having used variable shutter speeds before, how and in what way does the picture look less detailed at a shutter speed of 1/25 seconds compared to 1/50, other than in giving a lower *TEMPORAL* resolution in moving areas?

[Uppsala BildTeknik]

Sharpness and especially quality might be highly subjective terms (as interlacing lines might be regarded as a distortion, kind of blurredness, lack of visible detail)

Well if you are going to use a interlaced monitor to view the footage then the interlace fields are no problem at all, you will never see the fields.

If you are going to view the footage with a progressive monitor you should transfer the footage as progressive images to begin with, it would give you higher quality.

If you have interlaced footage full of visible interlace fields and you want to view it with a progressive monitor then you need to de-interlace it (it will look better). Perhaps there has been a misunderstanding from the beginning here, I thought you meant that by de-interlacing you gained better resolution, as in sharper, better images. (you don´t, unless you are comparing with interlaced footage full of visible interlace fields viewed with a progressive monitor...)

I wouldn't know if the above S8 examples were shot 'off the wall', they were transferred by Andec. But it doesn't matter, as I didn't post them to compete with the posted DV files, I posted them exclusively because they are S8 and this is what this forum is about, so their quality doesn't matter as they're not supposed to compete with the DV files here.

Or do you mean we shot the two DV files 'off the wall'?

I meant they are transferred "off the wall" as in recorded with a videocamera, off a wall, from a white paper, a mirror or whatever. Not with a frame-accurate transferring device. I was talking about the 8mm footage transferred to video.

I wonder what Andec used to transfer this...

That's why I gave you that link to those screenshots of my other project, to give you an idea of how sharp deinterlaced footage can be.

I didn´t look at it to be honest, because there was no "original" footage to compare with (prior to de-interlacing). You should compare a image from a interlaced film that has no visible interlace fields, de-interlace it and then compare the two images.

You lose quality by de-interlacing, you remove half of the image and replace it with a calculated "how it might be" kinda image. You lose quality even from parts of the image that have no moving things in them at all, you have to. You are removing 50% of the image.

Have you ever seen progressive images that need no de-interlacing? They are sharper. You should transfer a super8 reel with a frame-accurate transferring device and then apply a pulldown that does not introduce interlace artefacts (no mixing interlace fields of the footage). Then de-interlace it and see what happends to the quality. ;)

[mattias]

wow, this is probably the most confused thread ever on this board. almost all "facts" are off and the conclusions ever worse. no offense, discussion is a great way of learning, but i just thought i'd point it out so people don't take it too seriously. i won't enter the discussion but here are a few facts to set you on the right course: deinterlacing will make your footage both less smooth and less sharp, because you *are* dropping half the resolution temporally as well as spatially. just take my word for it.

/matt

[tlatosmd]

Sharpness and especially quality might be highly subjective terms (as interlacing lines might be regarded as a distortion, kind of blurredness, lack of visible detail)

Well if you are going to use a interlaced monitor to view the footage then the interlace fields are no problem at all, you will never see the fields.

I do, in the motions, it's probably the shutter speed.

Perhaps there has been a misunderstanding from the beginning here, I thought you meant that by de-interlacing you gained better resolution, as in sharper, better images. (you don´t, unless you are comparing with interlaced footage full of visible interlace fields viewed with a progressive monitor...)

Have you ever seen progressive images that need no de-interlacing? They are sharper.

I'm glad we agree on this. Material originating in progressive mode *does* look better than deinterlaced material, there's no question about it.

Um, sorry, I think I got it wrong about Andec though, it was Andec *processing* this Ektachrome roll, my co-student considered them also for transferring the stuff but it was too expensive and took too long for him so he decided on having a Hamburg-based company for it that would have one-day turnaround, would have to ask him about the name.

[Uppsala BildTeknik]

I do, in the motions, it's probably the shutter speed.

Impossible, definetly has nothing to do with the shutter speed. You cannot see the fields if the footage is properly transferred and interlaced, I wonder if the transfers are totally crap and that is why you can see the fields? Dunno, I have never actually seen or evaluated a crap-transfer so I don´t know.

If you could see the fields then you should see all the fields of footage shot with any videocamera also (they are all interlaced).

Anyone else care to come with some input, Mattias?

The filelds should not be visible because you do not see them at the same time, not with a interlaced monitor anyway. :?