A unique opportunity: results

Dan Margulis

I’ve spent a good part of this week completing my testing, and learning once again that there’s no end of learning to be done in our field. Here are the initial results.

To review: my vision is good, I don’t use glasses except for reading, but I am developing cataracts. I decided that the middle of a pandemic, when travel isn’t possible, is an ideal time to have them taken care of before they get worse. The first eye was done on 2 November and the second will be done tomorrow. The first eye has been stable all this week, meaning that my two eyes currently both see well and work well together, yet one sees colors quite differently than the other.

Consequently I temporarily have three different ways of examining an image, with the left eye alone, the right eye alone, and both together. The question was, do I adjust to each of these conditions and favor the same versions each time? It’s an important question to those of us who do this kind of thing for a living, because if our client doesn’t like our work, we would like to know whether it was 1) poor work on our part; 2) the client’s own personal taste; 3) the client happens to see color differently than we do.

The first can be found out by comparisons with the work of others, such as the case studies we recently went through. Several of us had the rueful experience of having our submissions demonstrated by others to be second-rate. But it is hard to know when the second or third variable is in play. Particularly, although we can certainly test a person’s visual acuity, we AFAIK haven’t got convincing proof up until now that it could ever make a difference, and if so, how often. And even if we knew that it could make a difference we still wouldn’t be able to quantify it because we couldn’t distinguish between preferences of personal taste and preferences caused by faulty perception.

That’s why my testing this week was unique. If I evaluated a certain set of images with one viewing method one day, and then another way several days later, well, my personal tastes are unchanged, so if I have a different opinion, the cause is perception.

The test suite consisted of evaluating 35 sets of at least four attempts to correct a given image. The source was archives of the best results from images that were at one time part of the Applied Color Theory class but were withdrawn, generally more than ten years ago. The advantages: I don’t remember specifics of how they were prepared, and all of the versions are pretty good. This is important because if a correction is seriously defective everyone will agree, even color-blind people.

I divided the sets into three groups, one each for the left eye, the right eye, and both together. When I had completed all 35, I divided each of the three groups in half and put them in line to be evaluated in a different way. When my 35 re-evaluations were done, I moved each set into the folder of the condition that had not yet evaluated it, and did 35 re-re-evaluations. I took notes of what I saw and what I preferred each time. At the end, I compared these notes and entered them into a spreadsheet to see what the differences were.

The sets covered a broad range of subjects:
*5 shots of people, faces moderately sized with respect to the entire image
*5 desert scenes
*5 other scenics
*4 interiors
*4 animal shots
*3 sports
*2 exteriors
*2 food
*2 people shots, faces large with respect to the image
*2 miscellaneous/unclassifiable
*1 people shot, faces small with respect to the image

I classified the results into four categories:
*8 sets had no significant differences in the three evaluations.
*15 sets had differences worth noting that may or may not have been caused by different perceptions.
*5 had significant differences in preference that were surely due to perception issues.
*7 had gross differences, such as one version being rated the winner in one evaluation yet as professionally unacceptable in another.

This was a much bigger impact than I thought, and I now understand, somewhat, the type of image where it comes into play. And because of it, I would now change my methods if I knew that the eventual client was an older person who had not had cataract surgery.

With a couple of exceptions the critical areas were always subdued reds and oranges, and bright yellows. Here’s why, I think:

The older eye sees through a filter of slightly warm yellow, which also reduces contrast. Consequently,

*Since everything is slightly yellow, the old eye is less sensitive to bright yellows. So it may accept yellow areas that the new eye thinks are too loud.

*To compensate, the old eye generally prefers a brighter, more vivid treatment of any image, EXCEPT if it involves subdued reds and oranges, such as are commonly found in fleshtones. Apparently, the old eye exaggerates these hues so as to better perceive them against the overall yellow cast. It therefore sees them as more saturated than the new eye, which is much more sensitive to blue, does. Example:

*The two large portraits were both of female models, one of African and one of Asian ancestry. In each case I picked a winner with the new eye, stating that the fleshtone was perfect. I rated those same two versions as unacceptably red with the old eye.

*An image of a stucco mansion, a slightly warm yellow. We happen to be given a reference version showing the color that the client prefers. Evaluating the contestants with the old eye, I disqualify one for being too orange, too far away from the reference color. With the new eye, the difference is far less pronounced and I consider the more orange one acceptable.

*A picture of a lynx kitten resting in a tree. My favorite version with the old eye is seen as unacceptably yellow by the new one.

I noted two lesser effects that are more understandable:

*Skies that tend toward cyan may be perceived as unacceptably green by the old eye.

*Where two images are basically equal, but one has marginally better sharpening or contrast control, the new eye will prefer that one, but the old eye may not detect the difference and will rate them as a tie.

I wondered whether the impact of using both eyes at once would be basically a 50-50 combination of the two. My guess, after looking at these results, is that the combination is more like a 70-30, with the good eye being favored for reasons known only to the forces of evolution. However I do not think this is an important point, since it only applies to people who have had cataract surgery on one eye but not the other.

If people want to discuss any of this further or see samples, I’m game; if not it’s still food for thought.


P.S. For obvious reasons, I won’t be participating in the group for a few days.