Slow motion of birds in flight?

[Barry Green]

Even though slow motion is best acheived with interlaced footage, for shooting fast paced movement that would not only be seen on regular TVs, I would still use the progressive mode, in order to get sharp picture like frames without interlacing artifacts.

You guys are missing Rob's point -- since the footage is going to be slowed down to slow motion, each interlaced field will be converted to a full frame. There won't be *any* interlace artifacts, there will be 50 frames per second.

If you want to capture as much motion as possible, you have to use interlaced mode. Progressive will capture 25 motion samples per second, interlaced will capture 50 per second. And after slowing down to 50% playback speed, there won't be any interlace artifacts, each field will be converted to a full frame and it'll look like film that was shot at 50 frames per second (although lower vertical resolution).

[Rob Lohman]

Okay, but how would you (seriously) do this with a video file?
Have you done it? In what program? The idea is that if you slow
footage down the program you do it in creates the inbetween
frames through a process known as interpolation. There are
various algorithms and thus varying degrees of succes. It sounds
like your describing something like this, although in an unusual
fashion (for me at least).

If you slow down interlaced footage 50% the program has half
a frame for every new frame and does not need to interpolate
the full frame (like it needs to do in progressive), but only half
with information from the surroundings.

And what Barry said (thanks!)... <g>

[Lauri Kettunen]

Rob, Barry, I think I understand your point, but ...

To minimize confusion, I put some sample frames (tif-files)available at

www.luontovideo.net/Birds.html

Now, could you download the Black woodpecker frames(progressive mode) and the Common kestrel or Siberian jay (interlaced mode) frames. That will give us some common ground to share our experiences.

If I understood Per properly, he wanted to have images like the Black woodpecker and show them in slow motion (say something like 2-4 frames per second).

[David Lach]

<<<-- Originally posted by Barry Green : You guys are missing Rob's point -- since the footage is going to be slowed down to slow motion, each interlaced field will be converted to a full frame. There won't be *any* interlace artifacts, there will be 50 frames per second.

If you want to capture as much motion as possible, you have to use interlaced mode. Progressive will capture 25 motion samples per second, interlaced will capture 50 per second. And after slowing down to 50% playback speed, there won't be any interlace artifacts, each field will be converted to a full frame and it'll look like film that was shot at 50 frames per second (although lower vertical resolution). -->>>

I'm aware that 50i will allow for more samples per second, but you will be losing half the resolution, which is a lot. When using progressive mode, you do not lose any resolution, but the extrapolated frames are farther apart. If you're only going to use the slow down feature in Premiere, the end result will look jerky. However, if you're using a morphing technique to create the extra in between frames, you'll keep both your resolution and smoothness of motion.

That being said, those morphing techniques have their limits and if the motion is too complex, the end result will look odd. Maybe not the best for bird shooting, as the movement might be too fast and therefore the difference between each frame too great, but it's worth a try. I always do sample tests when planing for slow motion and I try different techniques to decide which provides the best results. It's always a case by case decision.

[Per Johan Naesje]

Well, I think the discussion is moving towards technical aspects of video editing programs, plug-ins etc.
My initial question was: Settings on the camcorder to achieve the best “raw-film” result!

My biggest problem now, is that light conditions here in Northern part of Europe is hard, lots of bad weather, rain, snow…, Lauri: you agree with me?
Whish I could move, what about Hawaii?

- Per Johan

[Lauri Kettunen]

<<<-- is that light conditions here in Northern part of Europe is hard, lots of bad weather, rain, snow…, Lauri: you agree with me? ... - Per Johan -->>>

Luckily enough there was last night bright skies for couple hours and at the same time there was a magnificant Aurora borealis. Got the best footage (sharp and colorful, very little noise) I've ever managed to have Northern lights. So, I'm quite happy of being here right now, in fact!

[Rob Lohman]

<<<-- Originally posted by David Lach : I'm aware that 50i will allow for more samples per second, but you will be losing half the resolution, which is a lot. When using progressive mode, you do not lose any resolution, but the extrapolated frames are farther apart. -->>

This is true when you are just converting interlaced to progressive,
but that is NOT the case here. It is clearly stated that this is to
be used when ALSO DOING SLOWMOTION. Let me draw it out:

click here for difference between progressive and interlaced

Now if you convert this to 50% slowmotion it will be converted
to progressive. Look at the following:

click here for conversion to 50% slowmo

Now look at the difference. The resolution stays exactly the same
in both cases since all fields/frames are kept. However, the
information with interlaced is distributed across two frames instead
of one frame, thus each frame is constructed of less interpolated
material. Ofcourse these fields also have a slight temporal shift
which gives extra clarity on the moving object (and might even
allow you to get away with a bit lower shutter speed in case
you don't have a "enough" light, although not much).

> If you're only going to use the slow down feature in Premiere,
> the end result will look jerky. However, if you're using a

That might be, that depends on the algorithms they use. I can
imagine that it isn't geared towards a good end product for these
conditions like a specialized product does.

> morphing technique to create the extra in between frames,
> you'll keep both your resolution and smoothness of motion.

And how would one "activate" these "morphing techniques" in
a program like Premiere?

> That being said, those morphing techniques have their limits
> and if the motion is too complex, the end result will look odd.

ALL slow-motion algorithms that need to create inbetween
information (as is the case with our fixed rate camera's) use a
form of "morphing" (or as it is usually called: interpolation) to
create the missing information. The longer these gaps, the more
difficult and bad it will look.

> I always do sample tests when planing for slow motion and I
> try different techniques to decide which provides the best
> results. It's always a case by case decision.

That is definitely true and something everybody should be doing.
A technique that might work wonders one time might produce
bad results with some other footage and vice versa!

Just as a side note: we are talking about XL2 progressive mode
here which is a "true" progressive scan mode, not a faux one like
the XL1S. In that case different rules apply.

[David Lach]

click here for conversion to 50% slowmoNow look at the difference. The resolution stays exactly the same
in both cases since all fields/frames are kept. However, the
information with interlaced is distributed across two frames instead
of one frame, thus each frame is constructed of less interpolated
material. Ofcourse these fields also have a slight temporal shift
which gives extra clarity on the moving object (and might even
allow you to get away with a bit lower shutter speed in case
you don't have a "enough" light, although not much).

I think this is true on a theoritical level only, one I'm agreeing with from the start. It's obvious that you don't end up with more pixels captured by the XL2 over the entire footage using one method or the other. So in a way, yes, the resolution remains the same.

But, and your drawings there kind of point towards that direction, if you do a frame by frame analysis, you see that every slowed down frame in interlaced mode will only contain one half of the real world capture, hence my 50% lower resolution comment, whereas with the progressive scan, which on the PAL XL2 is the equivalent of taking a full frame photo every 25th of a second, you will have one frame out of two that will be entirely created from scratch and one full unaltered frame.

So unless I'm understanding this wrongly, and it is a definite possibility, the progressive mode, out of every 2 frames, will allow for one full unaltered 100% resolution frame and one in between frame created frome that full frame. The interlaced slow motion, on the other hand, will have each frame created with half the "reality" of what has been captured. Now of course I might not be understanding properly how the software recreates that other missing field for slowing down interlaced footage, but if it's from blending, you're definitelly losing resolution, where in progressive mode, there will be one frame out of two with no "captured reality" in it, but it will still have been created from one full resolution and unaltered image nonetheless.

And how would one "activate" these "morphing techniques" in
a program like Premiere?

Like I said in an earlier post, I use the Twixtor plugin from Revision Effects. Premiere alone does not do any kind of sophisticated morphing. Only frame blending, which isn't the same, or at least it isn't as advanced.

ALL slow-motion algorithms that need to create inbetween
information (as is the case with our fixed rate camera's) use a
form of "morphing" (or as it is usually called: interpolation) to
create the missing information. The longer these gaps, the more
difficult and bad it will look.

That's true. But what I call a morphing program is software that will not just blend the 2 frames together, but will look at the content (often you'll need to draw mattes around unmoving subjects to avoid background distortion), will calculate the motion vectors for literally each pixel and will draw the in between movements based on that analysis.

[Marty Hudzik]

I hate to jump in here so late but in general slowmotion created in editing software will first deintelace the footage to get 50% slow motion and if you slow it down more it will begin to just repeat frames to give the illusion of slower motion. I have been applying slow motion to wedding videos for 5years using all known versions of premiere and if you slow it more than 50% say to 33% you will see frames just blatantly being repeated. This could be considered slow motion.

But what I think we are all assuming is that the original poster was trying to achieve a "high speed" photogragphy look. Twixtor and Retimer are 2 plugins that will emulate this but to put it bluntly....they are expensive and a pain in the *** to use. It is not easy by any means.

There is just no affordable camera that will shoot 200-300 frames per second to achieve this effect. We have to cheat.

[David Lach]

<<<-- Originally posted by Marty Hudzik :

Twixtor and Retimer are 2 plugins that will emulate this but to put it bluntly....they are expensive and a pain in the *** to use. It is not easy by any means. -->>>

That's true. I wouldn't recommand using Twixtor if you don't have a LOT of time on your hands. I use it mainly for fictional stuff and it is very time consuming, but it can emulate the look of expensive high speed cameras when you know how to use it and most importantly, when to use it.

[Rob Lohman]

I think we are all sort of agreeing in the end. It all works or fails
with the quality of the algorithms being used and as we all
understand this can have a huge impact on how the reconstruction
is done and with what quality you end up with.

David: your points are certainly valid and are the same as mine,
the idea (and this has been tested) is that with a good slow-
motion algorithm (ie, not the Premiere included one) the inter-
laced method is usually giving smoother motion results (due
to the more even distribution of time). But as said earlier, test
, test, test and test <g>