Hutchinson C V, Ledgeway T, 2005, "Responses of first-order motion energy detectors to second-order images: Modeling artifacts and artifactual models" Perception 34 ECVP Abstract Supplement
Responses of first-order motion energy detectors to second-order images: Modeling artifacts and artifactual models
C V Hutchinson, T Ledgeway
Psychophysical studies (eg Smith and Ledgeway, 1997 Vision Research 37 45 - 62) suggest that contrast-modulated static noise patterns may be inadvertently contaminated by luminance artifacts (first-order motion) when carrier noise elements are relatively large owing to persistent clustering of elements with the same luminance polarity. However, previous studies that have modeled the responses of conventional motion-energy detectors to contrast-modulated static noise patterns have found no evidence of systematic first-order motion artifacts when the mean opponent-motion energy was used to quantify performance. In the present study, we sought to resolve this discrepancy. We subjected first-order (luminance modulations of either static or dynamic noise) and second-order (contrast modulations of either static or dynamic noise) patterns to conventional motion energy analysis. Using space - time representations of the stimuli we measured the net directional response to each motion pattern, using either the mean or the peak opponent-motion energies. As luminance artifacts that can arise in contrast-modulated static noise are predominantly local in nature, model responses were studied for a range (1 to 4 octaves) of spatial and temporal filter bandwidths. When the frequency bandwidth of the filters comprising the motion detectors was relatively broad (more localised in space - time) the peak (but not the mean) opponent-motion energy correctly predicts that detectable, local luminance artifacts are sometimes present in contrast-modulated noise patterns, but only when static noise carriers with relatively large elements are used. Furthermore, the model predicts that when dynamic noise carriers are employed (eg contrast-modulated dynamic noise and polarity-modulated dynamic noise), patterns remain artifact free. As such, the modeling and psychophysical results are readily reconciled. Our findings also demonstrate that the precise manner in which computational models of motion detection are implemented is crucial in determining their response to potential artifacts in second-order motion patterns.
These web-based abstracts are provided for ease of seaching and access, but certain aspects (such as as mathematics) may not appear in their optimum form. For the final published version of this abstract, please see
ECVP 2005 Abstract Supplement (complete) size: 2356 Kb