High Dynamic Range Video on two 5D's

[Lee Wilson]

And all I was saying is that Vegas allows sliding video by samples, if you disable quantize to frames.

I understand what you said about disabling "quantize to frames" (I think you already said that in an earlier post).

I am not sure I made myself clear (too many numbers!) - let me try and explain what I mean more clearly.

My point is that the idea that you can slide video by samples (or anything less than a whole frame) is merely a conceit within the application, you can't really do it with video footage.

After all what would the result look like (assuming we are blending two images as in a HDR application) - or what if we were simply editing between the two tracks with a simple cut, would track B come in a little earlier when we cut to it (if we have moved it earlier by a a few samples ?)

Of course track B would not (could not) come in earlier (even by just a few samples).

Take a look at this, I have moved track B earlier by less than a whole frame, now let's imagine the output from a 50% / 50% mix between the two tracks, what would you expect frame 147 to look like ?

Or if you were not blending two frames and simply doing a hard cut, what would you expect to see when cutting from frame 147 of track A to frame 148 of track B ?

 

[Petrus]

That shouldn´t be a problem at all with any NLE...

You guys do not seem to understand what is going on...

Every frame (pair) needes to be in PERFECT alignment and PERFECT timing for the system to work. If they are not, the resulting combined image will not be sharp.

For this reason only one lens (with beam splitter) can be used. For this reason each frame pair needs to be exposed the same instant (and same f-stop/shutter speed = ND filter must be used).

 

[Lee Wilson]

You guys do not seem to understand what is going on...

Every frame (pair) needes to be in PERFECT alignment and PERFECT timing for the system to work. If they are not, the resulting combined image will not be sharp.

In an ideal world you are quite right, although I think you can get away with some discrepancies - for example on a city shot (like the one in the example) where the camera is locked off - I think simply pressing the record buttons on the two cameras at the same time would be close enough.

For this reason only one lens (with beam splitter) can be used.

I can easily see a system using two lenses before the beam splitter. I can also see problems with convergence when sticking a beam splitter between the camera body and the lens (in the case of a single lens).

For this reason each frame pair needs to be exposed the same instant (and same f-stop/shutter speed = ND filter must be used).

You can also get around the need for an ND filter by employing a beam splitter that does the exposure difference for you - (for instance most good quality teleprompter glass splits the light as 70% transmission and 30% reflected).

I also don't think having identical shutter speeds is vital, but of course you don't want to mix 1/2000th with 1/50th as the result would be distracting. Same thing for f-stop, depending on your requirements there is no technical reason to have them identical.

 

[Steve_Karl]

My point is that the idea that you can slide video by samples (or anything less than a whole frame) is merely a conceit within the application, you can't really do it with video footage.

If all odd number frames had a red back ground and all even numbered frames had a lime green back ground ......
.... it depends on the content in the overlap.

 

[Steve_Karl]

You guys do not seem to understand what is going on...

Every frame (pair) needes to be in PERFECT alignment and PERFECT timing for the system to work. If they are not, the resulting combined image will not be sharp.

For this reason only one lens (with beam splitter) can be used. For this reason each frame pair needs to be exposed the same instant (and same f-stop/shutter speed = ND filter must be used).

Yes. That's understood now.
We just got off onto a tangent. Sorry.

 

[Lee Wilson]

If all odd number frames had a red back ground and all even numbered frames had a lime green back ground .......... it depends on the content in the overlap.

I don't know what you are saying here ?

Also I don't know why you are dividing the frames into odd numbers and even numbers ?

The content in the over lap will be: (assuming track B is shifted by 4,800 samples)

frame X of track A.

+ 90% of frame X of track B.

+ 10% of frame X+1 of track B.

The only way this could possibly be represented in the real world is to line up the tracks (no offset between A and B) and then blend the tracks using the percentages as above.

 

[Steve_Karl]

The overlapping area would be a blend of red and lime green but only for the distance of the offset and not the whole frame.

 

[Petrus]

In an ideal world you are quite right, although I think you can get away with some discrepancies - for example on a city shot (like the one in the example) where the camera is locked off - I think simply pressing the record buttons on the two cameras at the same time would be close enough.

I can easily see a system using two lenses before the beam splitter. I can also see problems with convergence when sticking a beam splitter between the camera body and the lens (in the case of a single lens).

You can also get around the need for an ND filter by employing a beam splitter that does the exposure difference for you - (for instance most good quality teleprompter glass splits the light as 70% transmission and 30% reflected).

I also don't think having identical shutter speeds is vital, but of course you don't want to mix 1/2000th with 1/50th as the result would be distracting. Same thing for f-stop, depending on your requirements there is no technical reason to have them identical.

The whole system is based on an idea of having ONE picture (or sequence of frames when using video) which is identical to just ONE picture, but combined form two differently exposed frames. If there are focus, depth of field, picture angle and point, timing or shutter speed differences between the frames, they WILL NOT form a perfect image when combined. Period.

Of course a beam splitter can split the rays in a ratio other than 50:50, that is trivial. That would work just as well as a ND. But the point is, exposure can not be varied by f-stop or shutter speed, or the moving images will not be identical in content (other than exposure).

The whole idea of this system is to achieve (a sort of) perfection. Why would you be ready to spoil it by deliberatly creating images which do not overlap perfectly, when it is possible.

 

[Lee Wilson]

The overlapping area would be a blend of red and lime green but only for the distance of the offset and not the whole frame.

There is no 'distance' to the offset - the offset is temporal, not spatial. By dragging something on a timeline to the left or to the right you are making it earlier or later in time - you are not positioning it spatially.

it sounds like you are suggesting this kind of thing . . . . (?)

Which would do this to my above example . . . .

Which brings us back to the original point - how is this making track B earlier in anyway ?

And if we have - for example - 10% of the left hand side of the next frame of track B appearing on the right of the current frame - what happens to the other 90% of the next frame of track B !?

C'mon, you see what I mean by now, there is no such thing as being able to offset video footage by anything less than a whole frame.

 

[Lee Wilson]

The whole system is based on an idea of having ONE picture (or sequence of frames when using video) which is identical to just ONE picture, but combined form two differently exposed frames. If there are focus, depth of field, picture angle and point, timing or shutter speed differences between the frames, they WILL NOT form a perfect image when combined. Period.

Thanks for the lesson.

:2vrolijk_08:

Like I say, you will be fine with some level of discrepancy, in fact with things like f-stop you might even like to play around with different values and experiment - there is no technical limitation here, and with things like shutter speed - on a locked off shot of a city for example - the difference between 1/100th for the highlights and 1/50th for the shadows would render such an imperceivable difference in motion blur as to be immaterial.

The whole idea of this system is to achieve (a sort of) perfection. Why would you be ready to spoil it by deliberatly creating images which do not overlap perfectly, when it is possible.

Ha ! A wonderfully emotive argument !

How can I possibly answer that, why am I deliberately setting out to spoil this 'perfection' ?

Hmmm . . . maybe I am just an evil person.

:cheesy:

Relax, experimenting with these parameters within reason will not break anything.

 

[Steve_Karl]

There is no 'distance' to the offset - the offset is temporal, not spatial. By dragging something on a timeline to the left or to the right you are making it earlier or later in time - you are not positioning it spatially.

it sounds like you are suggesting this kind of thing . . . . (?)

I was, and I tried it, and I see what you're saying now.
No need to hijack the thread any further.

 

[Lee Wilson]

I was, and I tried it, and I see what you're saying now.
No need to hijack the thread any further.

:thumbsup:

 

[Petrus]

Like I say, you will be fine with some level of discrepancy, in fact with things like f-stop you might even like to play around with different values and experiment - there is no technical limitation here, and with things like shutter speed - on a locked off shot of a city for example - the difference between 1/100th for the highlights and 1/50th for the shadows would render such an imperceivable difference in motion blur as to be immaterial.

The problem is, that large part of the midrange exposure values will be a combination of two frames, and also moving objects like cars in a cityscape would be combined from two images. I would find it odd if the highlights of a car were shot with short shutter speed and the shadow parts with long shutter speed, giving them motion blur. Same thing with playing with different apertures: more DOF on highlights, less on shadows of the same object...

Unless this is the funny effect your are striving for. For the intended use of this technique it is plainly wrong.

 

[Lee Wilson]

I would find it odd if the highlights of a car were shot with short shutter speed and the shadow parts with long shutter speed, giving them motion blur.

I doubt very much you would notice.

 

[Petrus]

OK, you can do your HDR videos any way you like as far as I am concerned...

If they turn out kind of funny looking, reread my posts.

 

[Lee Wilson]

OK, you can do your HDR videos any way you like as far as I am concerned...

Thanks for your permission, I wasn't waiting for it, but it's nice to have.

:2vrolijk_08: