Spatio-Temporal Masking: Hyperacuity and Local Adaptation

摘要

(a) Models of human vision were applied to image compression and image fidelityfor both static and dynamic images. The role of the human observer in JPEG compression was clarified. The human observer's visual information capacity was calculated to be about 20 bits/min2, substantially higher than previous estimates. New formulas were developed for the Fourier transform of JPEG basis functions, connecting JPEG quantization matrices to the human observer's contrast sensitivity function. Crawford masking was used to measure the visibility of lines and edges following abrupt luminance changes. The high frame rates produced higher temporal resolution than previous studies. A striking asymmetry between light and dark lines was found. (b) A robust test-pedestal framework was developed for modeling spatio-temporal vision with fewer assumptions than previous models. In this framework motion processing and hyperacuity thresholds are directly related to contrast processing. (c) A number of studies on motion processing developed new limits on the human visual system's capabilities. (d) In order to connect psychophysics results to underlying physiological mechanisms new techniques were developed for nonlinear analysis and source localization of visual evoked potentials and other biopotentials