Biol. Cybernetics 32, 9-17 (1979) Biological Cybernetics 9 by Springer-Verlag 1979 Human Psychophysics: Functional Interpretation for Contrast Sensitivity Versus Spatial Frequency Curve Allan W. Snyder 1 and Mandyam V. Srinivasan 1 f4 Institute of Advanced Studies, DePartments of Applied Mathematics and Neurobiology, Australian National University, Canberra, Australia Abstract. The hypothesis that neural processing in the human visual pathways compensates for both optical degradation as well as noise contamination at the photoreceptor level is introduced and shown to be consistent with the high frequency portion of the con- trast sensitivity function for threshold detection of si- nusoidal gratings in addition to the suprathreshold phenomenon of matching sinusoidal gratings of dif- ferent spatial frequencies. This offers a unifying inter- pretation for why, at threshold conditions, the high spatial frequency portion of the image is blurred as severely by the nervous system as it is by the optics (e.g. Campbell and Green, 1965) while in extreme supra- threshold conditions the nervous system effectively de- blurs the image (e.g. Georgeson and Sullivan, 1975; Kulikowski, 1976). These conclusions do not ne- cessitate a highly specific form of visual processing such as Fourier channeling. I. Introduction It is well known that the human visual system is more sensitive to certain spatial frequencies than to others (Schade, 1956; Campbell and Robson, 1968). Furthermore, the shape of the contrast sensitivity vs. spatial frequency function for threshold detection of sinusoidal gratings differs from an analogous function for suprathreshold conditions determined by matching sinusoidal gratings of different Spatial frequency (Watanabe et al., 1968; Blakemore et al., 1973; Georgeson and Sullivan, 1975 ; Kulikowski, 1976). For example, at threshold, sensitivity falls off rapidly above 1 This research was conducted at Yale University, Department of Ophthalmology and Visual Science, New Haven, Connecticut, USA, throughout which period A.W.S. was a John Simon Guggenheim fellow 5 c/deg while at suprathreshold the apparent contrast is comparatively invariant with spatial frequency. Here we show that the shape of the threshold contrast sensi- tivity function as well as the suprathreshold apparent contrast function is consistent with the hypothesis that neural processing compensates for optical degradation as well as noise contamination at the photoreceptor level. This offers one explanation for why, at threshold conditions, the high spatial frequency portion of the image is blurred as severely by the nervous system as by the optics in isolation. Our plan is first to determine the spatial transfer characteristics of a hypothetical visual system which forms a neural image most resembling the object world. Predictions based on this ideal system are then shown to be consistent with psychophysical measure- ments. We begin by considering the role of neural pro- cessing in extracting information about the visual en- vironment. This necessitates understanding the pro- pertie s of the input to the neural system, i.e., the image at the photoreceptor level. II. Attributes of the Image at the Photoreceptor Level Because of the deficiencies inherent in the eyes' optics (e.g. Campbell and Gubisch, 1966) the image at the photoreceptor level is lacking in its high spatial fre- quency content while even those lower frequencies which are present provide only a distorted version of the object (Campbell and Gubisch, 1966; Westheimer, 1972a). In other words the dioptrics behaves like a non- uniform low pass filter. Furthermore, this image is con- taminated by noise associated with photoabsorption and transduction (Barlow, 1964; Lamb and Simon, 1976; Yau, Lamb, Baylor, 1977). We now consider the properties that should be manifested by subsequent stages of neutral processing in order to extract the maximum information about the visual environment 0340-1200/79/0032/0009/S01.80