Dynamic ISO

[Patrick OMara]

Is there such a thing?

If you can detect over exposure, could you not then adjust ISO for only those regions.


Just curious/discussion.

 

[Phil Bates]

Actually, you can. For example, you can shoot at 800 ISO, and lets say you over expose the sky.

(This example will use After Effects.)
Import the clip and keep the ISO at 800. You can make a mask from the luminance of the image, isolating those blown out areas.
Then, duplicate the clip in the in the project window. Make that ISO 320. Then use the luminance mask as a traveling Matte to put the lower ISO version "inside" those overblown parts. Works great.

Phil

 

[Christopher Barrett]

I would love to see how this looks compared to HDRx.

Ugh, guess I'll hunker down this winter and start learning AE.

 

[Patrick OMara]

Yeah you could do it in after effects, good point. Although masking all that by hand could be tedious.

I wonder if for HDRX you could just capture those regions, leave out the rest for better compression, and for when you do a blend you aren't blending what you don't need to.

 

[Paul Nordin]

Hmmm, this seems like an unnecessarily complicated process. You can do the same thing by applying a curve to the raw image (LUT) which retains all the raw data without crushing highlight detail. HDRx on the other hand is using highlight info NOT in the baseline raw data, so some sort of masking is required to blend that into the image with your base exposure.

 

[Patrick OMara]

1. The curve - can you do that on cam? or you talking post. (i don't have an epic yet).

2. where does the "some sort of masking" come from?

 

[Keith Putnam]

Just remember that ISO is not "real"; it's only metadata affecting the brightness of the image on your monitor. You can't make the sensor more or less sensitive than it actually is. You can't change the gain, as it were, and you certainly can't selectively change the sensitivity of the sensor in only certain areas of the image.
What you have to do is ensure your highlights don't clip, and you do this by consulting the RAW exposure metering tools in the camera (or by employing HDRx, allowing the highlights to clip in your A stream because you'll blend in the X stream to replace them).
Capturing a properly exposed RAW image is the key. Once you've done that, achieving little to no clip in the highlights and staying above the noise floor in the shadows, you can manipulate selective areas of the image to your heart's content in post.

 

[Paul Nordin]

Keith said it exactly right.
Patrick, to answer your two questions:
1. the manipulations I'm referring to are in post. Curves and iso settings on camera only affect the monitor stream and metadate. What affects the image in camear (since its only capturing RAW and metadata) is Shutter Speed and Iris. PERIOD. Iso is just metadata and does not affect in any way what image information is recorded. You can find proper exposure to set your SS & Iris using what many think is the camera's native ISO of 320 (even number of stops from middle gray to black and white) or give yourself some leeway in protecting highlights by upping the ISO to 800 or more. But your setting the ISO on camera is not affecting what is recorded. Only SS and Iris, which physically adjust how much light hits the sensor do that.
Once the image has been captured and you are in Post, you can apply curves on the shot via any number of tools (AE, Davinci, RecCine Pro, etc.) to adjust where gamma lines up against highlights and blacks, and how much/little you want to crush your blacks/whites.

2. "Some sort of mask" refers to any post software that blends two images. In the case of HDRx, you are blending your A track (base exposed image) with your X track (highlight protection exposed image). you can use any standard and well known compositing techniques to do that, including the HDRx blending slider in Redcine X Pro.

I hope that helps

 

[Patrick OMara]

Well "in camera" you can know where you are over exposed. I am just wondering if there is any advantages to that considering it's all digital and there's a lot of new potential there, either to save time in post, or to help save on file size or possibly something completely new.

Whether it's helping with ISO, HDRx, or automatically creating a mask file, or sending the data to external hardware like a lighting system, or send to data a dynamic filter on your lense, or maybe to help with smaller file size for your x stream etc...

I dunno...

 

[Patrick OMara]

Thanks for the response Paul

Thinking about it, and you are right, anything you'd capture could be manipulated in post without too much trouble. Having the data in the file is always available to you and probably not worth making the camera do the work.

It could only ever be advantageous to things that happen in realtime while you are shooting, things that manipulate your environment to change the data that's coming in.

 

[Bob Gruen]

Info that truly clips is a voltage that overruns 1024, so 1024 is stored. There is a bit of a fudge factor built into the system, but best not to rely on it. Also the various functions involved are not linear.

Phil's method of masking in Ae would be fine as long as the hot items had not truly clipped, if they had (or even come close) then white is white and there is nothing to recover. This method may also result in banding or other artifacts.

There are several answer to the sky dilemma depending on your situation:

• Shoot the sky separately and do a replacement in Ae (see video copilot)
• Shoot with an ND Grad. filter
• Compromise your expose between the subject and sky (favoring the sky a little) and use Ae (or a finishing tool) to adjust separately.
• HDRx

HDRx is THE reason to own an Epic, period. Your need for over cranking in a film production can be predicted (and an Epic can be rented for that day), but when you are on location and a problem window is in your shot, simply engaging HDRx as a magic bullet solution will save you both time and frustration.

Bob

 

[Patrick OMara]

You could take the blown out pixels and use it as a mask for your xstream to only record the blown out areas. Should make for a better compressed xstream, ya?

 

[Phil Bates]

...You can do the same thing by applying a curve to the raw image (LUT) which retains all the raw data without crushing highlight detail...

Actually it's not the same, because a general curve will affect every part of the image, so darkening the sky will also darken other parts like the foreground. Patrick is asking for something that is region specific. By applying a levels adjustment to the luminance mask I referred to, you can isolate just the sky. In fact, better yet, use the blue channel instead of the luminance channel, making the process easier if the over exposed part is a specific color like a blue sky. I talked about this in a thread I started a while back and gave examples:

http://reduser.net/forum/showthread.php?46446-Tip-for-Making-the-Most-out-of-RED-s-Dynamic-Range-Outdoors&highlight=

Hmmm, this seems like an unnecessarily complicated process.

It doesn't have to be complicated. If you make a template with AE with this effect, it can be used with other clips simplifying the workflow.

Phil

 

[Phil Bates]

Phil's method of masking in Ae would be fine as long as the hot items had not truly clipped, if they had (or even come close) then white is white and there is nothing to recover. This method may also result in banding or other artifacts.

This is correct. I was referring to a sky that looked blown out when shown at 800 ISO only. My technique is for making the most out of multiple ISO's as long as nothing is truly clipped. Thanks for making that clear, Bob.

Phil

 

[Matthias Hutter]

Although I have never heard it stated specifically by Red, I believe the in camera ISO setting scales the sensor's voltage to the 10 bit storage registers at the analog to digital conversion. So when you adjust the camera's ISO you are adjusting what max voltage you expect to map to 1024 in the register..., ish..., kinda..., sorta...

Red doesn't use analog gain.

 

[Paul Nordin]

Actually it's not the same, because a general curve will affect every part of the image, so darkening the sky will also darken other parts like the foreground.

Yes, of course if you are creating an isolated adjustment. Sorry for mis-understanding your first repsonse.

Although I have never heard it stated specifically by Red, I believe the in camera ISO setting scales the sensor's voltage to the 10 bit storage registers at the analog to digital conversion.

Bob, I think you are wrong in this assumption. NO test I have seen or performed backs up this assertion. Every test I've seen, many comments by Jim Jannard, & company. and my own shooting experience fly in the face of your assertions. In practice, there is a native ISO that is ALWAYS what is recorded to the RAW files regardless of ISO setting. Anything else would fly in the face of the entire Red-Raw philosopy esposed by Red. ISO is just metadata. Don't confuse that with FUD.