Scratching beneath the surface of visual feature detectors
The notion of a feature detector has played and still plays a central role in thinking about human sensory processing. In its simplest form, this notion is embodied by the matched template model. I will discuss three lines of research that significantly extend/refine this model to account for detailed experimental measurements of the underlying perceptual mechanisms. The three topics are: 1) nonlinear processing; 2) internal noise; 3) natural statistics. When pieced together, these three additional modules expose a deeper structure to the notion of feature detectors, and offer a platform for framing potential directions of further research into their properties.