Abstract
A fundamental everyday task is visually searching for specific targets in the natural environments that surround us. In recent years, there have been a number of efforts directed at measuring and modeling detection of targets at specific locations in natural backgrounds, a key subtask of visual search. We have found that a productive approach is to bin natural background patches into joint histograms, with bins along background dimensions known from the classic vision literature to have a big impact on target detectability. By measuring psychometric functions in a sparse subset of these bins, it is possible to estimate how the included dimensions jointly affect detectability over the whole space of natural backgrounds. In previous studies, we found that human thresholds for sine and plaid targets are proportional to the separable product of the background luminance, contrast, similarity, and inverse partial-masking factor (a new dimension). Furthermore, this multidimensional Weber’s law was predicted by a simple template-matching observer with divisive normalization along each of the dimensions. The measure of similarity in these studies was the cosine similarity of the amplitude spectra of the target and background (SA)—a phase invariant measure of similarity. Here, we investigated the effect of the cosine similarity of target and background images (SI)—a phase dependent measure of similarity. We found that image similarity also has a substantial effect on threshold, and that threshold decreases monotonically with SI, in agreement with a recent study of visual search (Rideaux et al. JOV, 2022). Interestingly, the template-matching observer makes a completely different prediction: threshold is a U-shaped function of SI reaching a minimum when the target and background are orthogonal (SI = 0). Nevertheless, when the template-matching observer includes a small amount of intrinsic position uncertainty (which we measured in a separate experiment) the pattern of thresholds is predicted.