I have been astounded recently by the release of an application called "Lightbrush" - see:
http://www.tandent.com/support/
for tutorial videos and a user manual. The extraordinary aspect of Lightbrush is the application's ability to take a digital photograph and separate the tonal components (illumination, lighting data) from the color components (reflectance, surface information). This is a great big deal for CG artists that wish to derive texture maps from photo references, but need to get rid of lighting information in the image (shadows, for example) - they want nice flat maps that you can only get in overcast, flat or diffuse light.
What is extremely useful for digital photographers, or at least me, is the ability to independently edit the lighting and color data. It is mind boggling and I have yet to fully comprehend how powerful this approach will be, but here is a short example.
I shot this image as a horribly unappealing test of Canon-compatible Pocket Wizard mounted on a Fujifilm X100. Here is the result, with ACR7 defaults applied, to give you an idea of the lighting - hey, the flash fired, that was all I was interested in (there were actually two flashes, the main one firing from camera right). I was also doing this to see how I could compress the dynamic range of the scene (indoor-outdoor) with flash and still get a somewhat natural looking result. That came later in the test, but this crappy image has the perfect test elements for our exercise here.
Note the horrific shadows from the bare flash raking right-to-left across the chair back and onto the mug and writing desk, or whatever that fancy piece of furniture is called. Makes an unappealing image even more garish. This was a test to see if I could isolate that lighting information and then clone out some of the ugliness and come up with a convincing result.
So, I brought this image (raw file) into Lightbrush and went to work. THe user must instruct Lightbrush where "dark-light" pairs of surfaces exist to help it identify tonal data and separate it from reflectance data. Once done, you get the following kind of result:
The upper left is the original image, the upper right is the illumination data, the lower left is the reflectance data and the lower right is the composite - if you do no editing of either channel, the original and the composite are identical.
I have posted this example in the HDR forum because all of this magic is taking place in full 32bit land. You can send the result of the separation to PS for editing in full 32bit land - the tonal data of light and shadow can be manipulated independent of the reflectance data so that you can remove the shadows (either by boosting exposure/gamma or cloning nearby lighting values into the shadowed area) without destroying the underlying surface data. Pretty darn cool.
After about 10 minutes, here's what I came up with on my laptop using the trackpad:
If I took more time, I would have doused the specular hits on the wood a little bit and removed the shadow on the wall in the background as well. Next time!
What is also great is that after you've done all of your retouching, you can make standard adjustments to either layer - for example, adjust the gamma of the illumination layer to boost shadows and simultaneously add contrast to the reflectance layer. That's pretty awesome.
Well, I am pretty stunned at what this kind of approach can produce and, even in the hands of one not accustomed to working in such a way, the software is fairly straightforward to use. It is expensive, so I'm not sure what I will do at the end of the trial period.
kirk
