Humans spend most of their waking hours looking at regions filled with texture or pattern (regions of greater than 0% contrast) and very little time looking at blank un-patterned areas (0% contrast). And this visual contrast is changing rather rapidly in time if only as the result of eye movements. How does the contrast a moment ago affect the perception of contrast now?
A few years ago we discovered a rather surprising effect of short-term adaptation to visual contrast. If a test pattern is composed of two contrasts - one above and one below the contrast of the preceding adapt pattern - the test pattern is very difficult to perceive correctly. To explain this straddle effect, we proposed a rapid adaptation process in which a comparison level is continually updated at each spatial position. The comparison level equals the recent (less than a second) weighted average of contrast at that spatial position. The comparison level is subtracted from the current input contrast, and the magnitude of difference is sent upstream but information about the sign of that difference is lost or at least degraded.
We have more recently shown that this contrast-comparison process interacts with a more conventional contrast-gain- control process to produce Weber-law behavior both for contrast increments and for contrast decrements. Here we present a model composed of two parts: the contrast-comparison process that produces the straddle effect and a contrast-normalization network that - when driven by the output from the contrast-comparison process - produces the Weber-law behavior for both contrast increments and decrements. (A logarithmic transducer cannot explain this Weber- law behavior.) This model provides an excellent account of many psychophysical results including not only the overall effects but also individual differences in the irregularities of the results.
These experimental and theoretical results together suggest that the important quantity in much human contrast processing is not something monotonic with physical contrast. It is more like the (un-signed) difference between the current contrast and the recent average contrast.