Spatial Frequency Range in the Detection Process—1. Narrow Bars

Abstract

The effect upon the detectability of narrow bars of ideal low-pass and high-pass filters was investigated. The results indicate that for optimum detection of the stimulus, spatial frequency components up to about 10 c/deg need to be transmitted. Two models (a multi-channel model, incorporating probability summation across spatial frequency, and a Grating Sensitivity Curve model) were used in attempting to fit the results. The multi-channel model provided a more realistic fit, but both models ultimately failed because of non-linearity in the summation of responses to frequency components.

Spatial frequency range in the detection process—2. Wide bars and sharp edge

“…The fact that the multi-channel model fits the results for the triangular-bar test stimuli well is not surprising when one considers that the model has been derived from experiments in which test stimuli are best described in spatial frequency terms (Blakemore and Campbell, 1969;Graham, 1972;Sachs et al, 1971) and thus are confined to a narrow range of spatial frequencies. Both in Brogan (1985) and the present study it has been found that the model breaks down with a test stimulus whose Fourier transform extends over a relatively broad range of spatial frequencies. In Brogan (1985) it was postulated that (1) retinal imhomogeneity and/or (2) non-linearity in the summation of spatial frequency components, which were both omitted from the model, could explain this.…”

“…Both in Brogan (1985) and the present study it has been found that the model breaks down with a test stimulus whose Fourier transform extends over a relatively broad range of spatial frequencies. In Brogan (1985) it was postulated that (1) retinal imhomogeneity and/or (2) non-linearity in the summation of spatial frequency components, which were both omitted from the model, could explain this. It would seem that the latter would be more likely.…”

“…Since the Fourier transforms of the test stimuli in the present study are confined mainly to low spatial Spatial frequency range in the detection process-2 frequencies, it is not surprising that the results indicate that the range of spatial frequencies used in the detection of these stimuli is narrower than in the previous study (Brogan, 1985), in which the Fourier transforms of the test stimuli extended to relatively high spatial frequencies, and is confined to relatively low spatial frequencies. Consequently, the GSCs tend to have relatively narrow bandwidths and the slopes of the efficiency of detection curves are relatively large, with the results for the edge being an exception to this.…”

“…Consequently, the GSCs tend to have relatively narrow bandwidths and the slopes of the efficiency of detection curves are relatively large, with the results for the edge being an exception to this. As with the previous results (Brogan, 1985) the GSC model does not give an adequate description of the variation of efficiency with cut-off frequency. On the other hand, the multi-channel model does seem to fit the data quite well.…”

“…An earlier work (Brogan, 1985) proposed the technique of spatial-frequency filtering to gauge the range of spatial frequencies utilized in the detection of aperiodic stimuli, concentrating on test stimuli whose Fourier transforms extended over a wide range of spatial frequencies. Two models were used to fit the data: (1) a multi-channel model, incorporating probability summation across spatial frequency, and (2) a Grating Sensitivity Curve model.…”