Extracting Modulation Transfer Function of the Visual System from Contrast Sensitivity Function in External Noise
62.17, Wednesday, May 15, 10:45 am - 12:45 pm, Royal Ballroom 1-3
Zhong-Lin Lu1, Fang Hou1, Chang-Bing Huang2; 1Deparment of Psychology, The Ohio State University, 2Institute of Psychology, Chinese Academy of Sciences
The contrast sensitivity function (CSF) provides a comprehensive measure of spatial vision. Many have treated it as the spatial modulation transfer function (MTF), reflecting the gain of the visual system to various spatial frequencies, and used it to define the profile of the first stage filter in models of spatial vision (Ginsburg 1986; Chung, Legge and Tjan 2002; Watson and Ahumada 2005). Equating the MTF of the visual system to the CSF might however be problematic because contrast thresholds are jointly determined by the gain, internal noises and other nonlinearities in the visual system (Lu and Dosher 1999; Pelli and Farell 1999). Here, we applied the external noise paradigm and the perceptual template model (PTM) to separate contributions of the various factors to the CSF and extract the MTF of the visual system. Psychometric functions of five observers in a 10-AFC letter identification task were measured in five spatial frequencies and two external noise conditions (0 and 20%). The Sloan letter stimuli were filtered by a log-cosine filter centered at 3 cycles per letter with a 1-octave bandwidth, and scaled into different retinal spatial frequencies (0.5, 1, 2, 4 and 8 cpd). In the high external noise condition, the noise images were scaled with the letter size to maintain 18 noise elements per letter. We found that two factors contributed to the shape of the CSF: (1) additive noise accounted for 92±3 % of the variance, (2) a mostly flat but slightly dropped (about 11.8%) MTF at the highest spatial frequency accounted for an additional 4-5% of the variance. These results suggest that the MTF of the visual system, i.e., its gain to different spatial frequencies, is quite different from the CSF. The methods and results developed in this study may have broad implications for models of spatial vision.