Illumination Estimation Challenges

Photographer's opinion on the test submissions

In evaluating the submissions, I paid particular attention to the following characteristics and issues:

  1. Artefacts, particularly around light sources. There were very few.
  2. With a single main illuminant, neutralising any colour cast approximately, and avoiding a greenish or greenish-yellow cast.
  3. Less than full saturation of small bright colours.
  4. With the above in mind, a general at least moderate colourfulness.
  5. Detail visible in moderately dark areas while maintaining the very darkest as black.
  6. Prominent light sources (mainly street lamps) rendered with a balance of detail and flare — not too much of one or the other.

I also accepted that with the above taken care of, there will inevitably be a range of acceptable interpretation, particularly in the overall level of brightness, and there was indeed a range in the submissions from overall dark to overall bright. Colourfulness, too, is open to interpretation, although the submissions varied greatly, from very to almost monochromatic.

The scenes themselves varied in the problems they presented. Some, like IMG_1885, 8122, and 8133 were relatively easy. Others, like IMG_6139 (pedestrian crossing sign overexposed on lower half), IMG_7335 (massive contrast between lit central feature and surrounds), IMG_8087 (bright flaring light source close to important physical feature, the tower), and IMG_8131 (lighting difference between statues and the inscription when both need to be legible), predictably caused difficulties, and I paid particular attention to these when evaluating — there was great variation between submissions.

Signage, in fact, prompts the wider consideration of semantics in these scenes. Signage, people, activity are all salient features in photography, and could be (perhaps should be) enhanced. For example, IMG_8122 has both signage and people interest (in shadow) that I would have hoped would be raised in the rendering, but largely weren't.