More Neon Overexposure Issues

[Arthur Woo]

Some of you may remember my last test showed me the issues w/ blowing out neon lights and the issues that come with it. I read through various threads and tried using REDWideGamutRGB with Log310G and a color transform lut from Truecolor, as well as trying out the Lutify.me lut. No avail for this particular neon light. I'm a bit confused about what to do in these situations. For commercial and studio jobs, not a big deal. For exteriors for narrative and music video work, potentially a bigger issue as we steal locations from time to time, it would be a serious pain in the ass to have to continually have to watch out for lights for every shot. What do I do? Attached is an r3d example. As you can see, it's a night exterior. The ambient light is street lighting with sodium vapor, so the ISO is at 2000 and aperture is a T2.5, not unheard of for night exteriors w/ no film light fixtures. There's no way I can expose properly for the neon as well as the surrounding exterior location.

r3d for investigation: https://dl.dropboxusercontent.com/u/2999040/A001_C021_1215F3.0000206F.R3D

 

[Alberto Guglielmi]

Did you try with the new IPP2: http://www.reduser.net/forum/showthread.php?152780-Holiday-Update-IPP2-Preview-Build

 

[Arthur Woo]

I did not, thanks for the note. Is there no other way to fix existing footage? Currently in Resolve 12.5

 

[Dermot Shane]

Where's the issue? i'm not sure i'm seeing a huge problem, only a slightly clipped red channel;

red frame into Resolve,
ACEScct,
Acdemy highlight control DCTL in the media pool
red default settings other than iso, exposure, temp
709 ODT
contrast @ 435
exposure at -9
no other changes

 

[Marc Wielage]

Soft Clip and the Highlight control can help quite a bit. A lot of it boils down to a judgement call -- the key is to protect the highlights and add more fill. It's more about the fill, really. Without that, the pictures will get noisy.

 

[Arthur Woo]

Where's the issue? i'm not sure i'm seeing a huge problem, only a slightly clipped red channel;

red frame into Resolve,
ACEScct,
Acdemy highlight control DCTL in the media pool
red default settings other than iso, exposure, temp
709 ODT
contrast @ 435
exposure at -9
no other changes

I was hoping for more of something like what was posted in the IPP2 preview (see below). The sign doesn't look like that when you're just looking at it with your eyes. The red sign is a homogenous color with no hotspots. I noticed when I brought the exposure up, hot spots started to appear, so I intentionally clipped them to see what I could do in post. Neon lights have been an issue with Helium and I wanted to see if LUT/color space choices as pointed out in other threads could "fix" some of these issues. It's not a common occurrence, but say I'm shooting in the street and am exposing for the talent using available light. Protecting the highlights from neon signs would prevent me from doing this. Am I wrong to assume that Helium could act as a "doc" camera in such a situation, whereas I don't remember seeing this on Dragon and some other cameras?

 

[rand thompson]

Arthur,

With the preview Redcine-X Pro "One Off"(download below) if you chose "REC709" for Colorspace and "Linear" for Gamma Curve, you will get close to the "uniform red" of the sign that I believe you are trying to Achieve.


https://www.dropbox.com/s/p1zlrk5bzg85rzl/A001_C021_1215F3.0000206F_S000.0000000.jpg?dl=0


http://downloads.red.com/software/r...X_PRO_Build_42.0.41985_OneOff_For_Preview.dmg

http://downloads.red.com/software/r...uild_42.0.41985_OneOff_For_Preview_64-bit.msi

 

[rand thompson]

Just re-read your post, it shows when you lifted the exposure.

 

[Arthur Woo]

Hi Rand,

I updated Redcine-x but don't see "Linear" in the gamma curve options. I only see BT1886, HDR-2084, and LOG3G10. Am I missing something?

 

[rand thompson]

Arthur,

go under

1) Edit

2) Preferences

3) General

4) make sure "Show Advanced Controls" is checked

Make sure you are using the "Redcine-X Pro "One Off" version not the current version listed at Red.com. links below for Mac or Windows

http://downloads.red.com/software/r...X_PRO_Build_42.0.41985_OneOff_For_Preview.dmg

http://downloads.red.com/software/r...uild_42.0.41985_OneOff_For_Preview_64-bit.msi


You may also be clicking on the shortcut on your desktop for the version that you downloaded from red.com and not the shortcut that installed for the "One Off" preview version. The icon in my Windows 10 "Task bar" was still for the red.com version not the "One off" version

 

[Aaron Lochert]

I did not, thanks for the note. Is there no other way to fix existing footage? Currently in Resolve 12.5

Use RWG/Log3G10 + the IPP2 LUT. And then use an HSL qualifier on the highly saturated reds.

Desaturate a bit then compensate by pushing the offset wheel toward true red (watch the vectorscope for this). The net effect should be blending those reds/pinks into more of a monochromatic red tone.

 

[rand thompson]

Arthur,

Also, in the red.com version, you can use "Rec709" and anyone of the REDGAMMAs settings to get an even better look.


REC709-REDGAMMA 2 by rand thompson, on Flickr


REC709-REDGAMMA3 by rand thompson, on Flickr


REC709-REDGAMMA4 by rand thompson, on Flickr

 

[Aaron Lochert]

Rand, forgive me but those orange splotches still look pretty bad. You can smooth that out considerably going through the RWG/Log3G10 + IPP2 Preview.cube workflow.

Here's that workflow with a secondary on the reds.

 

[rand thompson]

You are right let me get rid of that.

REC709:color space and REDGAMMA 4: Gamma curve plus HSL secondary on yellow/orange color

[/url]Untitled00086400 by rand thompson, on Flickr[/IMG]

And Aaron that looks pretty good using your method.

 

[rand thompson]

Looked at Aaron's and my method of correcting Arthur's sign this morning and Aaron's Method of REDwidegamutRGB/Log3g10-IPP2 is the superior method.


My method of REC709 and REDGAMMA4 has more even "reds" but they are "way out of broadcast legality for red saturation.

[/url]REC709-REDGAMMA4 by rand thompson, on Flickr[/IMG]

Aaron's method with REDwidegamutRGB/Log3g10-IPP2 has much more legal red saturation


REDwidegamutRGB-Log3g10-IPP2 by rand thompson, on Flickr


And even if you tried to bring mine back into range, you would bring back the initial problem.

[/url]problem again by rand thompson, on Flickr[/IMG]

 

[Graeme Nattress]

Use RWG/Log3G10 + the IPP2 LUT. And then use an HSL qualifier on the highly saturated reds.

Desaturate a bit then compensate by pushing the offset wheel toward true red (watch the vectorscope for this). The net effect should be blending those reds/pinks into more of a monochromatic red tone.

So with gamut mapping there's a choice on what you can do when a colour goes out. Colours go out because they're too saturated, and because they're very bright and very saturated. For those colours that are very bright and very saturated they need to keep their brightness and loose their saturation.

If you take this shot under IPP2 and dial back the exposure, you see all the detail and saturation is there, but you'd need an HDR set to support that level of brightness AND saturation. So perhaps another way to grade this for SDR is to qualify on the red and grade it darker. That'll work to keep the saturated red and all the detail in there.

Graeme

 

[Hrvoje Simic]

If you take this shot under IPP2 and dial back the exposure, you see all the detail and saturation is there, but you'd need an HDR set to support that level of brightness AND saturation. So perhaps another way to grade this for SDR is to qualify on the red and grade it darker. That'll work to keep the saturated red and all the detail in there.

Graeme

And there is a third way.

 

[Dermot Shane]

If you take this shot under IPP2 and dial back the exposure, you see all the detail and saturation is there, but you'd need an HDR set to support that level of brightness AND saturation. So perhaps another way to grade this for SDR is to qualify on the red and grade it darker. That'll work to keep the saturated red and all the detail in there.

Graeme

in ACES wiht the ODT set to bypassed/none, this shot shows the red chanel being clipped, there would seem to be no detail to retrive in this case?

so varations on a theme of keying to fill the areas of clipped data with a color matching the rest of the sign has been the answer?

best answer might have been to stop down the lens by a 2/3 or so, there seems to be plenty of detail in the shadows and the highlights are forever clipped

 

[Hrvoje Simic]

Which highlights are forever clipped ?




(No qualifiers/masks/extractions used)

 

[Graeme Nattress]

"in ACES wiht the ODT set to bypassed/none, this shot shows the red chanel being clipped" - I don't think that's a valid way to determine clipped data. Looking at the raw data or my software version of the in-camera raw exposure check shows the red channel is not clipped. Looking at the brightest section I see raw 16bit code values of RAW 51749,14596,1236 (16) which is indeed very hot, but not actually clipped.

The real problem is that the red light is both very bright and very saturated. Very bright very saturated colours don't fit nicely in RGB colour spaces unless you greatly expand the brightness axis (as in go to HDR). Think of it this way - take an RGB cube and stand it on black, with white at the highest point. The most saturated colours - pure red, green or blue are now half the peak height of the cube. The peak height of the cube is pure what. What that means is any bright colour (with a brightness approaching that of pure white) must be increasingly desaturated the brighter it gets.

Graeme