Noise-based characterization of small-scale algorithms for visual processing
Recent advances in visual psychophysics have permitted a more detailed characterization of basic visual processes such as bar/edge identification and stereoscopic surface detection. As a result, the potential link to corresponding phenomena in single neurons has been made more transparent and available for highly constrained modelling using small-scale algorithms, i.e. simple plausible models that use a small set of neuronal operators. I will discuss both published and unpublished material that demonstrates how this approach is not only useful for bridging physiology and psychophysics, but also for establishing conceptual links across different visual systems (human versus insect) and different levels of visual processing (low-level versus higher-level). Published material will be presented briefly, to allow time for a more detailed discussion of unpublished data from a recent project on joint encoding of motion and disparity signals in human vision.