Highlights from the HDR Symposium

Stanford's HDR Symposium turned out to be very interesting event. The collective brainpower of a small nation, crammed into a single room - you bet the air was sizzling from ideas and strong opinions.



Highslide JS



Here's a quick recap of some selected points of interest:


  • Marc Levoy explained how the Camera 2.0 project will enable a community-driven approach to push computational photography forward. Key element is the departure from "black box" firmware, in favor of a fully scriptable open-source platform.
    Levoy further recommends Animal Eyes as reference book for everyone building capturing devices.

  • Helge Seetzen from Dolby Labs shared some interesting insights on the difficulties of driving Local Dimming hardware. He called color LEDs "little buggers" for being notoriously inconsistent in color. But exploiting their flaws rather than fighting them leads ultimately to an even better display. For example, spectral leakage turns into an advantage when driving 6 instead of 3 primary colors, resulting in a much wider gamut.

  • Jack Tumblin lets us rethink what we consider a "great image", and explains how current tonemapping methods might be missing the point. He honored Renaissance artists like Rembrandt as excellent tonemappers, using artistic liberty to cheat the lighting to create more evolving images.
    Tumblin recognizes the "evocative HDR Look" for inducing an emotional response, but the result is often achieved by muddling through. In this regard he proposes a tonemapping approach, that looks beyond the pixels on the screen, and rather makes a distinction between surface colors, reflection, and lighting. This would allow more controlled look-finding. For example, you could tweak just the lighting in a photo, without worrying about side effects like "dirty" or "super-glossy".

  • James Ferwerda is hitting the same vein by proposing an extension of the La*b* color space with variables describing the glossiness of a material: c* for contrast gloss and d* for distinctness-of-image.
    In a perfect execution of the scientific method he did a field study where participants rated the gloss impression of a rendered ball, which resulted in a psychovisual gloss model. Although originally geared towards CG imaging, this model could be interesting when applied to photography in general. Also, Ferwerda dropped a new wording that I find very appropriate: High-Fidelity Imaging. Yay.

There was a lot more going on, and I will most certainly get into detail in later posts. HDRI is a pretty wide field, and this HDR Symposioum surely succeeded in bringing the top guns from adjacent fields together. Kudos to Joyce Farrell for flawlessly organizing this remarkable event.

Just one question remains unanswered:
Who owns tonemapping? The camera, the photo software, or the display device?

Next Page