High Dynamic Range Imaging at SIGGRAPH 2003

At SIGGRAPH 2003 in San Diego I searched for developments in high dynamic range imagery ( HDRI ) and the techniques associated with it.
I found lots of expert information in several venues at this year’s conference. In addition to courses and sketches on the subject I found new products that create and present HDR imagery in Emerging Technologies and Exhibits.

Take a picture, scan it, capture it, and transfer it to your computer for enhancement with your preferred software package. Lately, the results displayed on your computer monitor seem more than satisfactory given today’s advances in computer graphics technology. However, if you look around in the real world with normal human eyes you see a much larger range of colors and luminance values. Our vision capabilities process contrast ratios of 2,000,000:1 and see a much broader color spectrum than the mere millions of colors available on most presentation devices. Your graphics card might be high end, but without the right display device it can’t quite represent a reality.

At SIGGRAPH I witnessed some stunning image displays that have again raised the bar on image quality. It was in the Emerging Technologies venue that Sunnybrook Technologies presented the HDR Display. Capable of representing the high dynamic range imagery that has started a buzz at the last few SIGGRAPH conventions, the HDR display is 30 times brighter than any commercially available display technology while producing a black that is 10 times darker. They have combined an LED panel for adding precise luminance values behind an LCD panel for color values. The results give impressive improvements. They say the ratio is 60,000:1 from the darkest to the lightest portion of the screen. Compare this to the 600:1 contrast ratio LCD monitors that are offered currently.

For some years now, Paul Debevec, www.debevec.org, has been the featured speaker at the SIGGRAPH conference on topics such as HDRI, Image-based Lighting, and Light probes. He has presented very useful production techniques and demystified methods for merging real world imagery with computer graphics. He has also been featured in the Electronic Theater four times if you didn’t already know.

In the SIGGRAPH 2003 course, HDRI and Image-based Lighting, Debevec shared the platform with Greg Ward, Industrial Light & Magic’s Rod Bogart, Weta Digital’s Dan Lemmon, and Frank Vitz, an independent VFX supervisor. They covered topics on HDR, light acquisition, and compositing for film while emphasizing real world production pipelines. This class was packed as they covered ground on the making of X-Men2, Debevec’s Fiat Lux , and Rendering with Natural Light.

Some familiarity of cameras and photography is useful to understand how images are processed. When you expose your camera’s film to higher apertures ( f-stop ) this restricts the amount of light that strikes the film and thus the shadows in the picture show less detail, but the lighter areas tend to look right. The opposite technique of letting more light in adds detail to the darker recesses of the picture while washing out the highlights. Digital cameras give similar results.

If you can capture multiple exposures of the same shot using a tripod to lock down the camera and combine all these together to make one image then you can control more precisely how the light and dark areas can be manipulated. Brightness and contrast can be adjusted across the whole image without losing any detail. The same thing our eyes do when we enter a dark place (go to your dark place now and try it out).

If you are interested in making your own high dynamic range images try HDRShop to process a set of multiple exposed images into one HDR formatted image. http://www.ict.usc.edu/graphics/HDRShop/download . The resulting images can be used in several 3D software packages for rendering and texturing as well as high end digital compositing software that can take advantage of 32 bit floating point data.

On the Exhibits floor there were cameras that automate the process. Spheron showed a camera that could create panoramic images in HDR format in one step. See information here, http://www.spheron.com/info/SpheroCam_HDR.pdf

A discussion in the SIGGRAPH Sketches venue added some facets to the HDR genre. High Dynamic Range and Tone Mapping, presented by a panel of experts showed some more techniques where calculation could remove shadows from an image ( assumptions are that a sequence could easily be batch processed ) allowing you to create new shadows and change their direction to whatever suits your needs! Maybe in the future it won’t matter what time of day you shoot your scenes you can go back and correct shadows or correct lighting mistakes in post. That’s where everything is fixed, right?

At the same sketch ILM introduced some open source image format that they say meets flexible needs for cinematic effects. As well as operating in 32 bit floating point, it allows you to imbed normal direction and other CG information. It even tells you how each image is represented in six degrees of Kevin Bacon.
http://www.openexr.com/