Spatio-Temporal Masking: Hyperacuity and Local Adaptation

摘要

Our development of an ideal-observer framework and a test-pedestal methodologyfor modeling vision without the numerous assumptions of previous models has provided a comprehensive understanding of the spatio-temporal characteristics of human vision. The methodology encompasses a limited set of test stimuli with a multiplicity of pedestals to facilitate the comparison of performance across many psychophysical tasks. For example, it is shown that vernier acuity can generally be predicted from an individual's contrast discrimination threshold. For the conditions under which contrast discrimination predictions break down, a detailed modeling of later stages of visual processing is required. As a result, specifications for a vision modeling tool have been developed to guide the creation of a comprehensive vision modeling environment. As our models of visual function have matured, we have applied them to practical issues such as image compression and image quality. Consideration of properties of human vision is essential if the image compression needed for new technologies such as HDTV are to avoid sacrificing image quality. The success of the test-pedestal methodology has also lead us to record human visual evoked potentials so that we may integrate our psychophysical data and models of vision with underlying physiological mechanisms.