C100: Subtle PsF Interlacing Artifacts

W

wgtwo

Well-known member
What's up with this subtle interlacing left behind when viewing (properly interpreted) PsF footage? This stuff: http://i.imgur.com/UZAS53E.jpg

I may be pixel peeping, but it's as plain as day to me, especially when looking at, say, people's faces from far away. Fine details in a wider shot. Did some tests, and it does not appear to be confined to a single colour channel, as some posts I have read seem to imply.

I've read that article that tells you to use 5DtoRGB, and that doesn't help at all.

Am I doing something wrong, or is this just normal? Anything that can be done about it?

Also, on an unrelated note: is it possible to conform 50i down to 25p without losing detail (so it is the same as 25PsF), in order to have the possibility of making slo-mo footage in post? I believe it should be, but couldn't figure out how.
 
wgtwo said:
Also, on an unrelated note: is it possible to conform 50i down to 25p without losing detail (so it is the same as 25PsF), in order to have the possibility of making slo-mo footage in post? I believe it should be, but couldn't figure out how.
Click to expand...

No. Recording 50i gives you two time separated fields per frame, whereas 25p gives you whole frames and 25PsF gives you a whole frames split into two non time separated fields.
 
AndreeOnline said:
No. Recording 50i gives you two time separated fields per frame, whereas 25p gives you whole frames and 25PsF gives you a whole frames split into two non time separated fields.
Click to expand...
Thought so, thanks. Is this the case with all interlaced recordings? Haven't had much in-depth experience with interlaced cameras before.
 
wgtwo said:
Thought so, thanks. Is this the case with all interlaced recordings?
Click to expand...

Yes.

That said, due to its non debayerd, crisp image to start with, quite a few on the forum have been happy recording 50i or 60i and then using that as a base for slow motion.

Moving the camera, or filming something in motion, usually introduces motion blur on either a frame or subject level. That makes the loss of actual resolution less noticeable.
 
Yeah, I'm happy enough with using 50i for slow motion, just not a huge fan of having to specifically shoot shots as slow motion shots. This is from coming from a GH3, with which I could just shoot a whole event in 60p and then slow a small portion of my footage down in post. Oh well, I can live with that.

Anyone got anything to say about the PsF interlacing artifacts?
 
Tell us more about the image. I looked at it, but it's hard to make anything out since it's just a small crop.

I see a very dark image with a hint of maybe a shirt collar and the highlight of an ear? Provide more context and/or a larger image with more obvious signs of what is going wrong....
 
Take a close look and you can see lines with resembling interlacing artifacts, particularly around the ear. Just cropped tight to emphasize it.
 
Zooming in, I do see a faint horizontal pattern. I don't really see it at 100%. It firmly falls within pixel peeping territory, but there's nothing wrong with that when you want to understand your footage better.

To me, there doesn't seem to be any pixel shift between the two fields, meaning the footage is correctly de-interlaced. I don't know if this is an actual thing happening in the compression stage in the camera with very small differences in luminosity between fields, or if it is a post thing that is software dependant.

What software do you use (apart from 5DtoRGB) and what is your workflow?

While it's interesting to know what it is, I definitely wouldn't worry about it. Just for context—provide the whole frame at 100%.
 
wgtwo said:
Yeah, I'm happy enough with using 50i for slow motion, just not a huge fan of having to specifically shoot shots as slow motion shots. This is from coming from a GH3, with which I could just shoot a whole event in 60p and then slow a small portion of my footage down in post. Oh well, I can live with that.

Anyone got anything to say about the PsF interlacing artifacts?
Click to expand...

Mmm, you can do that. Just shoot 50i and if your NLE is up to the task, you'll just edit in a 25p timeline and when you slow down the NLE will automagically use the available 50i fields.

Of course, if you use a s*%tty NLE like FCP7, then it won't work.

And give us a better image, it's hard to see if those are just regular interlacing lines that come from movement (remember that the 50i image is displayed both fields at the same time when watching on a computer without any processing) or something more weird.
 
Oh yeah, should have clarified, it is as Andree said - no pixel shift, just luma variance. Not ACTUALLY interlacing, but I strongly suspect it has something to do with the PsF container. It resembles interlacing in that it is a perfect on/off line by line pattern; fields 1, 3, 5 are one colour and 2, 4, 6 are another etc.

Currently rendering something, but I will provide a full frame tomorrow. I'm surprised others haven't noticed it - I remember looking at some sample footage before buying the camera and seeing it.

Edit: Here we are - http://i.imgur.com/iwSgDT2.jpg Again, it's definitely not movement related pixel shift (though it looks like it might be in some portions of this shot). JPEG compression breaks it up a bit, but it's still quite visible.
 
Last edited:
wgtwo said:
Not ACTUALLY interlacing, but I strongly suspect it has something to do with the PsF container.
Click to expand...

I've seen this.

From my investigations, I'd strongly wager that it's partly compression artefacts, partly the strange thing that the Canon does in making a PsF. If you dual-record onto a separate I-frame recorder, it will be devoid of such artefacts.

Now, if you really pixel-peep deep (!), try flicking through the frames. Note that there's a cadance to them which is the IBBP cadance of the AVCHD compressor. There's a good frame, a couple of stripey bad ones, then a 'better but not quite best' frame, couple more bad ones, and so on until you get the whole Long-GOP cadance: 15 frames for NTSC area rates, 12 frames for PAL area rates.

To my eye, it's at its worst when you're in bland or low contrast lighting, using gain, and haven't done an ABB in a while. Test shots indoors using domestic lighting when you're not going the whole hog with setups usually generate the worst results. LOL
 
I find that I can see it in I frames too. It's a little less obvious because of the finer grain structure, though.

And yeah, from what I've seen it only really manifests itself in low light. However, that is exactly my concern - what happens if I am shooting in low light, as I often do?

Either way, web compression should break it up enough, and I imagine I'd be using a recorder for anything not intended for just the web. Still, I'm interested to know what exactly is causing it.
 
You must log in or register to reply here.
Back
Top