The central area of the primate retina contains two types of cone photoreceptors: L (red) cones, sensitive to long wavelenghs of light, and M (green) cones, sensitive to middle wavelengths of light. The use of a multiple cone type receptor array presents two fundamental problems regarding the encoding of spatial and spectral information. First, a uniform luminance signal no longer produces a uniform response. Second, spectral information about the visual scene is not as complete as luminance information because each location in the visual scene is sampled by only one cone type.
To test specific hypotheses regarding how the retina encodes both spectral and spatial information, I have constructed an ideal observer (Green and Swets, 1974) for a retinal model that incorporates photon noise and all known pre-neural and neural factors up the level of the cone synaptic terminal. Using simulated psychophysical tasks, the ideal observer performance at a given retinal stage characterizes the information loss and processing at that level and allows easy comparison to human psychophysical performance. By adding additional stages of the anatomically defined retinal circuitry to the ideal observer, we can begin to understand the encoding scheme used by the retina and how optimal the retina functions as a sensor for the visual system.
Back to Andrew's Home Page