A Statistical Perspective.

This paper applies statistical design principles to a simple biological model of human vision so that we can more clearly interpret the apparent role of eye saccades. In doing so we show that many structural features of the biological system (such as the optical geometry of the retina) are not complications which need to be surmounted or accomodated by later processing, but very good strategies for minimising the resources required to construct a working image recognition system. Therefore, (and as expected) the structure we see can very easily be explained by evolutionary pressures. The work also suggests specific forms for intermediate computed quantities, chosen to match the representation of signals as pulse sequences in neuronal systems. The ideas presented have implications for the construction of artificial (computer) vision systems. The computational model is very closely related to (but not based upon) SIFT, but more strongly based on a consideration of vision as a process of measurment while also linking the idea of multi-scale analysis with biological structure