Abstract

When one asks about the properties of visual signals stimulating nervous systems, the ultimate interest lies in determining how the signal is encoded and transferred across the receptor barrier, how much and what kind of information is passed further to the higher visual centers and what is lost. The research presented in this dissertation attempts to explore quantitatively some detailed aspects of information processing by microvillar photoreceptors.

Until recently, three methods were used to measure or estimate information transfer from the visual scene to the photoreceptor: Shannon’s information capacity, the closely related linear coherence rate and the compression entropy rate. In the first research article of this dissertation, a novel information calculation method based on principal component analysis, the mutual information rate, was developed. In the second publication, the influence of a physiological delay in the photoreceptor response on the information rate estimates by the Shannon method-related coherence rate algorithm was explored and a technique to compensate the associated error was proposed.

The third study addresses the question of whether photoreceptors can more efficiently transfer information arriving from natural sources than from common artificial visual stimuli. Natural stimuli have interesting statistical properties in the form of higher order correlations (HOC), arising from the presence of features representing surfaces, textures, and object boundaries. This problem was investigated in the most extensive study to date, using blowfly Calliphora vicina photoreceptors as a model. The individual photoreceptors encode input information by a form of Weber’s law, with the HOC in natural sequences reducing information transfer by decreasing the number of local contrast events that exceed the noise-imposed threshold.