Grayscale Calculation of Foot Candles

[Ken Schafer]

Here's our issue -- we're working on a software program for cinematographers that uses 3D rendered lighting. The light sources in the 3D engine do follow the inverse square law but use their own arbitrary brightness units with no equivalancy factor between their units and foot candles (or lumens, lux or any other real-world measurement for that matter).

 

Now, I realize that that unlike the real world, in the software we can only perceive the 3D world through a virtual camera (which by definition will have an f/stop and "speed") so I know that I have to make some arbitrary choices.

 

However, given a chosen f/stop and film speed, I believe I should be able to back-calculate it if I could get images of a calibrated grayscale card lit solely by a light source with known foot candle specs. In other words, if someone could

 

1. Take a single-bulb fresnel light with published foot candle specs (such as all Mole-Richardson's lights)
   
2. Light a calibrated grayscale card solely by that light
   
3. Shoot images of the grayscale card with the light at the distances listed by the lights foot candle specs
   
4. Keep the camera fixed and use a single f/stop for all images (and yes, I know this will probably result in some images being under or over-exposed)
   

Given the above images and camera distance, f/stop and film speed, I believe I should be able to sample the resulting images and recreate it in our virtual world so that the virtual grayscale card matches the actually shot images, thus calibrating our software's internal lights.

 

 

In case that wasn't clear, for the Mole-Richardson's baby 1K shown above, we'd need six images of the grayscale card, one each at 5, 10 and 15 feet in full flood, and again in full spot. We'd also need to know the film speed, f/stop and camera distance from the card.

 

Does this approach seem reasonable, and are there such images already available?

 

If it does seem reasonable but such images do not already exist, if someone would help us by generating them, we'd be certainly willing to give free software (worth several hundred dollars) and credit.

 

Any takers?

[Jonathan Bowerbank]

Do you want it to be shot on video or do you want it done on film, like an emulsion test?

 

I can easily do it for you, if it'll help you in developing your software.

 

Video, I can do it for free, no problem.

 

Film, I'll have to charge for stock/processing/telecine :)

[Hal Smith]

Hopefully you're aware that any extended source does not fall off by inverse square? You have to integrate a constellation of point sources (each following inverse square) over the area of the source to predict falloff. A 6" fresnel behaves pretty much inverse square in the far field but there is a near field range where it's not following inverse square. You'll need to deal with extended sources since so many lights in current use are extended either directly (Kinos, etc.) or by use of diffusers in front of the light (Chimera's, etc.). And then there's my favorite approach, bouncing light off ceilings, walls, cards, (actors fannies? :) ) etc.

 

Are you familiar with the concept of LUT's (Look Up Tables) in monitors/displays/projectors/video cameras? It seems what you're saying you need it an LUT to transform between the physical world and film/video response. Another topic where we can mourn John Pytlak's death, he'd be all over this topic like a glove.

[Guest Glen Alexander]

Here's our issue -- we're working on a software program for cinematographers that uses 3D rendered lighting. The light sources in the 3D engine do follow the inverse square law but use their own arbitrary brightness units with no equivalancy factor between their units and foot candles (or lumens, lux or any other real-world measurement for that matter).

 

Now, I realize that that unlike the real world, in the software we can only perceive the 3D world through a virtual camera (which by definition will have an f/stop and "speed") so I know that I have to make some arbitrary choices.

 

However, given a chosen f/stop and film speed, I believe I should be able to back-calculate it if I could get images of a calibrated grayscale card lit solely by a light source with known foot candle specs. In other words, if someone could

 

1. Take a single-bulb fresnel light with published foot candle specs (such as all Mole-Richardson's lights)
   
2. Light a calibrated grayscale card solely by that light
   
3. Shoot images of the grayscale card with the light at the distances listed by the lights foot candle specs
   
4. Keep the camera fixed and use a single f/stop for all images (and yes, I know this will probably result in some images being under or over-exposed)
   

Given the above images and camera distance, f/stop and film speed, I believe I should be able to sample the resulting images and recreate it in our virtual world so that the virtual grayscale card matches the actually shot images, thus calibrating our software's internal lights.

 

 

In case that wasn't clear, for the Mole-Richardson's baby 1K shown above, we'd need six images of the grayscale card, one each at 5, 10 and 15 feet in full flood, and again in full spot. We'd also need to know the film speed, f/stop and camera distance from the card.

 

Does this approach seem reasonable, and are there such images already available?

 

If it does seem reasonable but such images do not already exist, if someone would help us by generating them, we'd be certainly willing to give free software (worth several hundred dollars) and credit.

 

Any takers?

 

Is this like Maya or similar? I toyed around with those packages and don't like them because they have no modeling ability of the physics in the 'real'world. I want to be able to know the differential equations or integrals used, not just look pretty.

 

If you want to see how well you are doing your similations, look at a real world program like Zemax, it will be a way to have a reference. There are other optical programs, but look for a good ray tracer. Autodesk used to make a 'lighting' program that modeled real world electrical lighting for streets, signs, etc. have a google.

 

For your current post, you need to know the optical curve of your light source with the lens, the material of the board. If you try to model anything else with Maya that has different optical properties, there will be errors, i.e., if the refence board is made of cardboard vs plastic, each has different absorption, scattering, re-radiation and reflection coefficients for the wavelengths of light. You would almost need a database of optical properties that you're interested in utilizing versus wavelength.

 

When you add in the characteristics of the human eye, it gets complex, so also depends upon how much accuracy your solution requires. If you want a first order, very simple linear solution or some 5rth order partial differential equation. ha ha The overall specs you've listed are insufficient, i.e., not enough information to properly formulate an answer. Not being negative, just honest IMHO.

 

If you want a good and free calibration paper for black and white, i suggest you have a look at this website

 

http://research.microsoft.com/~zhang/

 

I've read and used zhang's research papers and algorithms professionally for years. Zhang knows his stuff, no doubt, he's a vision guru IMHO you'll probably be interested in this paper

 

3D Dynamic Scene Analysis: A Stereo Based Approach (with O. Faugeras) (Springer, Berlin, Heidelberg, 1992). ISBN 3-540-55429-7 & ISBN 0-387-55429-7

 

under his Projects, check out this

 

Camera calibration

 

it's a gold mine.

[Ken Schafer]

Do you want it to be shot on video or do you want it done on film, like an emulsion test?

 

I can easily do it for you, if it'll help you in developing your software.

 

Video, I can do it for free, no problem.

 

Film, I'll have to charge for stock/processing/telecine :)

 

Hey Jonathan:

 

Thanks for the kind offer; but actually Chris Nibley beat you to it. He was going to be calibrating a new light meter anyway and so we're doing a two-for-one. Depending how it goes, I hope I can keep your offer "open" as I may want to come back to you with other questions or testing if you're willing.

[Ken Schafer]

Hopefully you're aware that any extended source does not fall off by inverse square? You have to integrate a constellation of point sources (each following inverse square) over the area of the source to predict falloff. A 6" fresnel behaves pretty much inverse square in the far field but there is a near field range where it's not following inverse square. You'll need to deal with extended sources since so many lights in current use are extended either directly (Kinos, etc.) or by use of diffusers in front of the light (Chimera's, etc.). And then there's my favorite approach, bouncing light off ceilings, walls, cards, (actors fannies? :) ) etc.

 

Are you familiar with the concept of LUT's (Look Up Tables) in monitors/displays/projectors/video cameras? It seems what you're saying you need it an LUT to transform between the physical world and film/video response. Another topic where we can mourn John Pytlak's death, he'd be all over this topic like a glove.

 

I do and unfortunately, for our purposes, we can't take any of it into account because we're not doing Ray Tracing, just real-time point and spotlight lighting with true DOF and multi-light source shadows. However, that means no diffusion nor bounced light. I know, I know, you're all thinking so what's the point... :)

 

The answer is that the program is not designed to be a lighting plan designer, but rather is real-time previsualization and presentation. However, within those known and admitted lighting limitations, I'd like to get as close to the real-world output as possible, so if someone wants to put a 1K of known capabilities on a set, it will throw roughly the same amount of light as the real-world fixture.

[Ken Schafer]

Is this like Maya or similar? I toyed around with those packages and don't like them because they have no modeling ability of the physics in the 'real'world.

 

Hey Glen, thanks so much for the info, references and thought that went into your response. I feel like I went to use a water fountain and got a fire hose!

 

While our program and Maya have some loose similarities (they're both powerful 3D programs with cameras) they have very different focuses. Maya is a general purpose 3D modeling/rendering program while ours is a dedicated 3D previsualization program that uses a large library of stock objects and characters (though you can build and import your own) with built-in relationships and interactions.

 

Ours is designed specifically for filmmakers and is user friendly enough that pretty much anyone with basic computer skills can use it in very short order. It offers real-time previsualization with detailed set and camera information reports. While we are integrating basic lighting and shadows, it will not however, be a lighting plan designer as it is real-time and doesn't use any diffusion, light bouncing or other advanced techniques that require ray tracing.

 

I looked at Zemax and that's orders of magnitude more complicated than what we're looking to do, which is a quick and rough light calibration so that a light like the Mole Richardson above, throws roughly as much light in our 3D world as its real-world counterpart. I know it will be rough, but without ANY calibration between our lighting numbers and the real world, I feel like the guy in Spinal Tap who keeps on insisting that "It goes up to ELEVEN!"

[Jonathan Bowerbank]

Thanks for the kind offer; but actually Chris Nibley beat you to it. He was going to be calibrating a new light meter anyway and so we're doing a two-for-one. Depending how it goes, I hope I can keep your offer "open" as I may want to come back to you with other questions or testing if you're willing.

 

Right on, a good guy to get that test done for ya :)

 

Feel free to get in touch with me if I could be of any service. I've been interested in seeing a quality software for set lighting. WYSIWYG seems to have cornered the market right now, but I found it quite counter-intuitive. So do update us on your progress as well!

[Hal Smith]

WYSIWYG seems to have cornered the market right now, but I found it quite counter-intuitive.

 

All of the high end lighting programs are pretty arcane. I've got LD Assistant for Autocad and it's got a pretty steep learning curve but at least I was able to leverage my 15 years of experience with Autocad. I had WYSIWYG and got so frustrated with its crippled drafting functions that I sold it on eBay. I kept looking for basic drafting functions like trim, osnap, etc. and they weren't there. Even Autocad's el cheapo imitators include that level of functionality.

[Guest Glen Alexander]

All of the high end lighting programs are pretty arcane. I've got LD Assistant for Autocad and it's got a pretty steep learning curve but at least I was able to leverage my 15 years of experience with Autocad. I had WYSIWYG and got so frustrated with its crippled drafting functions that I sold it on eBay. I kept looking for basic drafting functions like trim, osnap, etc. and they weren't there. Even Autocad's el cheapo imitators include that level of functionality.

 

Hal,

 

Even with Inventor 2k8, I still keep my ancient copy of AcadLite at the go. Yes it's old, dxf, but brutually fast for anythng simple and 2D. I'll have my drawings done in AcadLite before Inventor finishes loading. :)

[Guest Glen Alexander]

Hi Ken,

When I go into the studio with the musos, I always make that joke,"It's goes to ELEVEN." ha ha I tried Maya and thought ok here's a teapot, here's a brick, let's smash them into each.... like that Bugs Bunny cartoon with Marvin the Martian, "Where's the Ka-Boom? Where's the earth shattering Ka-Boom?" ... meh..

[Ken Schafer]

Right on, a good guy to get that test done for ya :)

 

Feel free to get in touch with me if I could be of any service. I've been interested in seeing a quality software for set lighting. WYSIWYG seems to have cornered the market right now, but I found it quite counter-intuitive. So do update us on your progress as well!

 

I will. We are toying with the idea of outputing to POV Ray or another similar Ray Tracer which would theoretically render all the nuances of the lighting plan, including reflections, bounced light and so on. How much use would this be to you folks? What would be the minimum features a lighting plan designer would need for it to be useful to a DP?

[Brad Dickson]

Having worked in WYSIWYG, LD Assistant, 3D Studio, ESP Vision and have been shown Rosco's software, Martin's software, along with a host of other pre viz software not one of them does it all well. Having looked at your product video I doubt people are going to have the time to wait on radiosity renderings. If you have a very fast rendering system and can simulate going through various types of diffusion etc. you may want to license it to other software developers.

However I suspect to keep your price low and to keep it all real time you'll best to keep it simple and stay with basic sources. Keeping an up to date library of fixtures with their photometrics is what makes the difference in lighting software.

I personally find the pre viz portion of the software to be a general idea to convey a concept and not to be photo accurate. I have a friend who does 3D work for the Royal Opera House in London. For stage work it is a real time saver but for film with so many light sources being soft a real time renderer that requires a basic computer to render may be hard to accomplish. However the other software rendering programs don't cater strongly to film so if you accomplish this it would be good.

Edited January 22, 2008 by Brad Dickson

[Ken Schafer]

Having worked in WYSIWYG, LD Assistant, 3D Studio, ESP Vision and have been shown Rosco's software, Martin's software, along with a host of other pre viz software not one of them does it all well. Having looked at your product video I doubt people are going to have the time to wait on radiosity renderings. If you have a very fast rendering system and can simulate going through various types of diffusion etc. you may want to license it to other software developers.

However I suspect to keep your price low and to keep it all real time you'll best to keep it simple and stay with basic sources. Keeping an up to date library of fixtures with their photometrics is what makes the difference in lighting software.

I personally find the pre viz portion of the software to be a general idea to convey a concept and not to be photo accurate. I have a friend who does 3D work for the Royal Opera House in London. For stage work it is a real time saver but for film with so many light sources being soft a real time renderer that requires a basic computer to render may be hard to accomplish. However the other software rendering programs don't cater strongly to film so if you accomplish this it would be good.

 

Thanks for the feedback.