Abstract
Maximally camouflaged animals mimic the luminance, contrast, colour and texture of their background, so that the only cue left usable for detection is the edge. This is one of the hardest cases of visual detection, so studying this reveals many of our detection strategies and their limits. We conduct psychophysical experiments where humans detect a diverse array of synthetic camouflage targets of varying textures and shapes. In parallel, we develop a theoretical and computational detection model that is meant to closely emulate the human visual computations for this task. The model first filters an image with the human contrast sensitivity function, then uses biologically plausible mechanisms to extract edge gradients and group them into contours, both at the true object boundary, and the edge contours of the texture which may mask the true boundary. Next, it suppresses the weak edges and computes several features of the remaining contours, such as their length, edge power at various scales, and curvature. It then measures how conspicuous are the features of the boundary contours compared to the texture contours, to produce an overall detection response. We fit these to our experimental data so that the single principled model can predict human camouflage detection performance across our entire array of diverse stimuli, and account for many of our parametric experimental observations.