Exposure Calculation with Photoshop?

[Garrett Shannon]

I was curious if anybody knew of a program or a way to do the following in Photoshop or something similar:

 

I would like to be able to look at a digital still image on my computer and know what the different levels of exposure are within the image. Preferably in stops or something easily translatable to stops ie: the left side of the face is 1 2/3 stops hotter than the right side.

 

Imagine having a still image from a movie and mousing over various parts of the image to get a general idea of how contrasty faces are lit or if there is a back light how many stops over it is. I think it would be a useful tool both to study the work of other DP's as well as review and make notes on my own work.

 

Let me know if this is confusing and I can try to clarify. Otherwise, thanks. =)

[K Borowski]

I was curious if anybody knew of a program or a way to do the following in Photoshop or something similar:

 

I would like to be able to look at a digital still image on my computer and know what the different levels of exposure are within the image. Preferably in stops or something easily translatable to stops ie: the left side of the face is 1 2/3 stops hotter than the right side.

 

Imagine having a still image from a movie and mousing over various parts of the image to get a general idea of how contrasty faces are lit or if there is a back light how many stops over it is. I think it would be a useful tool both to study the work of other DP's as well as review and make notes on my own work.

 

Let me know if this is confusing and I can try to clarify. Otherwise, thanks. =)

 

I hesitate at fraternizing with one of the enemy, but here goes. . . ;)

 

You need to shoot grey cards or test charts at each lighting-level area within the frame, then just measure them with the eyedropper tool. This will give you a definite, exact, easy reading of the relative levels of exposure.

 

Of course, you need to have the damned grey cards at the right angle though, so something very very analog is still at the heart of the accuracy of something very very digital :P

 

Actually, in seriousness though, this is actually no different than what has been done for decades, centuries even with densitometry readings. You also have the added flexibility of working with white or black, though I'd avoid black.

[Tim Terner]

To guess a ball park figure about what is over and what is underin a shot isn't that difficult, but to do it your way in Photoshop would also need the 'dialling in' of the stocks used

[Garrett Shannon]

I don't need to know what the exact exposure is, only the relative exposure between one area of the image and another. Is there some way to translate luminance values in photoshop to stops so i can know for example that one area in the image is 1.5 stops brighter than another area in the image?

[K Borowski]

I don't need to know what the exact exposure is, only the relative exposure between one area of the image and another. Is there some way to translate luminance values in photoshop to stops so i can know for example that one area in the image is 1.5 stops brighter than another area in the image?

 

Sorry Garrett, I don't know for certain, but I *think* I heard somewhere that 10 points is comparable to a stop. I'm sure someone on here knows. . .

[Scott Fritzshall]

It's kind of tricky because most of what you're looking at is in sRGB colorspace, in which brightness scales nonlinearly. It's also just a range from black to white, rather than measuring exposure. Beyond that, you've got no idea what the original image looked like before it was squeezed into its current colorspace. There could be two highlights that have 3 stops difference between them, but you've got no way to tell because in the 8bit image you're looking at, they're both just "white." You'll have slightly better luck if you can look at it in a linear colorpace where 1 is twice the brightness of .5, but I'm not sure how Photoshop handles that, and the aforementioned problem is still there- you may be looking at heavily-modified values in the first place.

[K Borowski]

Thanks for chiming in Scott. This isn't my area of expertise. Now that I think of it, maybe it wasn't photoshop with the 10-point scale. Do you know of any color-correction programs designed for quick color and density correction that do use a linear, or a camera-stop derived scale?

[Chris Keth]

To do it, you would just have to shoot a test. Light a scene by monitor and then, from the camera, take note of what everything reads with a spotmeter.

 

If you do this, another useful things to do is to evenly light a 21 step greyscale. That will show you how the camera progresses from black to white and how far under key blocks p and how far over key blows out.

[Tim Terner]

To do it, you would just have to shoot a test. Light a scene by monitor and then, from the camera, take note of what everything reads with a spotmeter.

 

If you do this, another useful things to do is to evenly light a 21 step greyscale. That will show you how the camera progresses from black to white and how far under key blocks p and how far over key blows out.

 

Something like this http://www.motion.kodak.com/motion/uploade...ve_data5217.pdf

[Garrett Shannon]

Something like this http://www.motion.kodak.com/motion/uploade...ve_data5217.pdf

 

 

Ya. That is exactly what I am looking to be able to do. Just on any image.

 

Is this what I would need to do the test that Chris Keth talked about: http://www.stouffer.net/Reflection.htm ?

 

Also, how do I look at an image in a linear color space? Forgive my ignorance but I don't know that much color space and technical digital image lingo. Thanks everyone!

[Scott Fritzshall]

Thanks for chiming in Scott. This isn't my area of expertise. Now that I think of it, maybe it wasn't photoshop with the 10-point scale. Do you know of any color-correction programs designed for quick color and density correction that do use a linear, or a camera-stop derived scale?

Nuke converts everything to 32bit floating point, so rather than seeing values from 0-255, you'll see them from 0-1. If you bring in an sRGB image and tell Nuke to treat it as linear, then it removes the sRGB gamma curve and remaps the brightness so that .5 is twice as bright as .25. I just remembered, actually, that it has a handy little function in the viewer where you can have it give you "spot meter" readings from the pixel values, so that you can actually see how bright something is. This works regardless of whether your image is sRGB or linear.

[Scott Fritzshall]

Ya. That is exactly what I am looking to be able to do. Just on any image.

I really don't think you can. The meter readings really don't line up with the pixel values in that image. That's the whole point of having different stocks- that specific values of light will translate to different densities. Plus then the negative gets scanned or transferred, and they stuff all of that exposure data into just 8 bits and manipulate the contrast and whatnot. It's possible that I'm wrong, and that someone else can explain why, but as far as I can see, there are about a dozen different factors that play into what kind of pixel values you get at the receiving end, and you don't know the parameters for any of them.

 

Also, how do I look at an image in a linear color space? Forgive my ignorance but I don't know that much color space and technical digital image lingo. Thanks everyone!

I'm not exactly sure how to do this in Photoshop. What you can do is set the gamma to 2.2, which will cancel out the sRGB gamma curve and make it basically linear, but I don't think that this is mathematically identical to viewing it in linear colorspace so I'm not sure if it's accurate.

 

sRGB is a computer thing designed to display images the way we see them. Our eyes don't see color linearly, so if you map colors in an 8bit image so that 4 is twice as bright as 2, 8 is twice as bright as 4 (this would be considered mathematically linear), etc, you get something that looks very flat and washed out. sRGB remaps the colors so that they are perceptually linear, and thus look like what we would expect to see. Just about everything you see on your computer is in sRGB colorspace, unless you're dealing with HDRs, which are mathematically linear so that they can store exposure data.

[Scott Fritzshall]

Thanks for chiming in Scott. This isn't my area of expertise. Now that I think of it, maybe it wasn't photoshop with the 10-point scale. Do you know of any color-correction programs designed for quick color and density correction that do use a linear, or a camera-stop derived scale?

Actually I just remembered that Photoshop has some neat features in the color picker window if you set the image to 32bit. You can take any color value and ask for it to be any arbitrary number of stops brighter or darker. It gives you samples of what it would look like +/- 3 stops too. It's pretty cool.

[Joshua Csehak]

Why not just shoot a gray card wide open, at the point where it's just blown out, snap a pic, stop down, snap a pic, and keep going till it comes out black, and however many pictures you've got, that's the dynamic range?

[Chris Keth]

Why not just shoot a gray card wide open, at the point where it's just blown out, snap a pic, stop down, snap a pic, and keep going till it comes out black, and however many pictures you've got, that's the dynamic range?

 

Because if you can do it in one image, why do it in more?

[Joshua Csehak]

Because if you can do it in one image, why do it in more?

 

I guess if you set up a scene with objects of varying levels of brightness and spot-metered each one and determined that they each were a stop brighter than the previous one, and there were enough of them, you could take a picture of the scene and see what got blown out and count down till it got too noisy for you. I think the multi-pic method is way easier, though.

 

Unless, are you talking about determining a camera's DR based on a scene you haven't metered? I don't see how that's possible.