Focus Assist

[Jim Arthurs]

It might go without saying, but I found nothing either in the RED store under EVF or searching under EVF diopter to confirm this. Nor do the photos of the EVF show a focusing ring to indicate thus.

At NAB I remember twisting the diopter to my eyesight. I remember it being rather tight and sticky, and I had to go to the limit for my vision in one direction.

Unless I was dreaming.

Heck, it's possible I dreamed up everything, like going to NAB and seeing an affordable 4K camera.

As I mentioned months ago and others have brought up right here, I again put my vote towards the histogram thingy visable on playback as well. Really the bane of the current crop of HD cameras, not being about to recheck your focus on the built in EVF/LCD as well as you can check when live and recording...

 

[I Bloom]

The Focus level dictates how severe the transparent focus overlay is, at 30% it gives a general fuzziness to the overlay around the focussed areas with the most in focus areas being warmer. With the focus level at 100% it picks out only the 0.08% (Made up figure) that the camera estimates is 100% in focus.

Paul, what you want is exactly what I want but I see some difficulties. For example it seems like just using edge detection would give you a good idea of whats in focus for example I can redux you image below in photoshop and use the following steps.

1. Use Edge detection on the image.
2. Select a theshhold of the darkest pixels from the edge detection using color range.
3. Invert the selection and clear everything else.
4. Invert the selection and fill in red.
5. Overlay it on the original.

I get something like below:

And that seems like it's showing whats in focus. SEEMS LIKE. But it doesn't take into account contrast. A high contrast out of focus edge will also trigger the edge threshhold, like for example look at the edge of the pen below.

It's insidious because it seems to be working over there on her shoulder (my focus was off as you can seeinch but on the pen the technique is clearly wrong.

This is, I think what Graeme is talking about when he says:

I personally found the JVC system misleadingly telling me something was in focus when it was not, and the coloured edge (which in the HDV camera I used was not as thick as the edge in the link above) obscured the edge you were trying to see!

I've been considering how you can solve this problem, not that I'm the guy too ask and I think what it would involve might be a Fourier Transform of the image to find high freqencies in the image followed by an inverse fourier transform to overlay those areas. That would be nearly if not more expensive computationally than REDCODE. This is all my guess.

This is also what I think Graeme means when he says:

Paul, would be fabulous if your diagram could be done via an algorithm, but I doubt it's easily possible, especially in the type of software that can run in real time in camera.

The idea behind the waveform is that you can see the height of the graph and know whether something is in or just partially in focus. This does help as then you can judge degree of focus.

So the question then becomes for me what is the waveform then, because what we normally think of as waveforms seems like it could readily translate back into an overlay. Well maybe it's a lot easier to compute a one dimensional transform than a two dimensional one, and also easier to not reverse-transform it back.

So my guess is that what you will see is a stripped down transform of each vertical line maybe on just the green channel. This appears as a waveform but each mark on the graph doesn't correspond to a pixel, it corresponds to a frequency. It basically tells you if that frequency is present in the image on that vertical line. Again I'm really guessing I don't feel like I can talk about this with confidence.

So Graeme my question: is there a way we can roughly localize that information on the image. Say within a block of pixels. So that one intuitively can see that the right area of the image includes some high frequency data but not exactly where? My fear is of having to look away from the image to get the info I need and of a possibly difficult comparison between the image and the waveform.

As below:

Thanks,
IBloom

 

[Brook Willard]

At NAB I remember twisting the diopter to my eyesight. I remember it being rather tight and sticky, and I had to go to the limit for my vision in one direction.

Unless I was dreaming.

Heck, it's possible I dreamed up everything, like going to NAB and seeing an affordable 4K camera.

As I mentioned months ago and others have brought up right here, I again put my vote towards the histogram thingy visable on playback as well. Really the bane of the current crop of HD cameras, not being about to recheck your focus on the built in EVF/LCD as well as you can check when live and recording...

The EVF definitely had an adjustable diopter - I used it myself.

 

[Finner]

I also played with the EVF diaopter and talked to stuart about how a numbered bezel around the diaopter would be good so the 1st AC could remember the director is 9.5, himself and the camera op are 9, and the DOP is 10 for example. That way each person does not have to find the sharp setting every time it is changed and just spin it to there correct # setting. I sure hope this was done.

 

[Brook Willard]

That makes two of us... I'd hate to have to put camera tape around the EVF.