Overexposure

#1

I was debating which forum to post this in, and I opted for this one, because I’ll present the concept from a programming perspective.

I was thinking about the subject of overexposure within the concept of a full 3d rendered scene (in particular, a scene rendered with a RenderMan compliant renderer), and whether the traditional pipeline allows for it. I’ve been reading about Cineon file formats (10 bit log) and how it allows extra headroom for overexposed pixels. However, I have yet to find anything which really addresses the concept of ‘bleeding’.

Visualize an interior shot exposed for the interior lighting, and a window to the exterior showing an overexposed outdoor environment. Film will tend to bleed, where the brightness of the overexposed exterior bleeds to the frame around the window on the inside.

From an algorithm standpoint, it seems fairly simple as a post render step: keep scanning the pixels in the image for pixels which exceed the display capability of the final output device. For those pixels which meet that criteria, apply a filter (box, gaussian, etc.) to spread the brightness outward. Keep iterating over the pixels in the image until there are no pixels which need to spread (‘bleed’) their brightness.

Now, some questions: is the above described technique a standard applicable step in any of the compositing programs, such as Shake or After Effects? If not, is the technique often applied somewhere else in a pipeline, and at what stage, and through what software? Or, is another technique applied instead?

#2

This process is traditionally applied in Shake (or whatever compositing package you like) as a post-process. You should render to exr since it’s got the best dynamic range.

The simplest way to get the effect you’re describing, bloom, is simply to add a blurred copy of the original image back over itself. The blurred copy should be much darker so the bloom is only apparent where the original image is very bright. If you want to get more fancy you can blur each channel different amounts to fudge chromatic abberations.

Ultimately the best way to do this is to use a filter that approximates the effect of the lens “properly”. I’ve seen a good implementaion that does something funky with transforming the image into frequency space and doing some kind of convolution there, but that’s as much as I know about the technique I’m afraid.

#3

Thank you for the reply. The various ways one could manage the process through color channel transformations, etc. prior to the blur make it clear that you could have a fair degree of fine control by using one of the compositing programs.

#4

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.