| This dissertation investigated how correlated (filtered) noise affects efficiency and human strategies for performing two visual pattern recognition tasks: Gabor patch orientation discrimination and three-dimensional object recognition. Compared with the ideal observer, human observers were most efficient at performing these tasks in low-pass noise, followed by white noise; they were least efficient in high-pass noise. Contrast-dependent internal noise appeared to limit human performance in the high-pass conditions. Classification images did not predict the observed pattern of efficiencies in white and low-pass noise, and so effects of internal noise and nonlinear strategies were investigated. In the Gabor patch task, humans had lower internal noise in the low-pass conditions than in the white noise condition, whereas internal noise levels were similar across conditions in the object recognition task. Simulations demonstrated that the same nonlinear strategy could have different effects on efficiency depending on the noise condition, but also suggested that the amounts and types of nonlinearities in the human strategies were likely to have differed across conditions. Taken together, the results of this dissertation demonstrate that not only efficiencies, but also levels of internal noise and the linearity of the human strategy for performing a visual pattern recognition task can be influenced by the correlation structure of the external noise. |
