Christopher DiMattina¹ (), Curtis L. Baker Jr.²; ¹Florida Gulf Coast University, ²McGill University

Edges in natural images often give rise to spatially correlated changes in both first-order visual features (e.g., luminance), and second-order features (e.g., contrast or texture). Given the co-occurrence of first- and second-order cues at edges, it is important to understand how they are integrated for purposes of edge detection, and whether different kinds of second-order cues interact in a different manner with first-order luminance cues. Here, we compare the interactions between luminance modulation (LM) and contrast modulation (CM) with the interactions between LM and orientation modulation (OM) in an edge-detection task. Human observers detect which of two temporal intervals contain an edge between halves of a disc-shaped region with a filtered noise "texture" or carrier, whose contrast, orientation, or luminance differ across the boundary. We first characterize dependence on carrier spatial frequency, demonstrating different "tuning curves" for CM and OM, neither of which is explained by the carrier contrast sensitivity function. We then perform sub-threshold summation experiments in which observers detect either CM or OM in a 2IFC task with an LM pedestal varying from sub-threshold to supra-threshold levels. We find that an LM pedestal does not facilitate CM detection at sub-threshold levels, and actually masks CM at supra-threshold levels, consistent with previous studies. However we find that an LM pedestal has little or no effect on OM detection. When we consider detection of LM with CM and OM pedestals, CM pedestals impair the detection of LM, whereas OM pedestals do not. Therefore, we see that different forms of second-order structure interact in a different manner with first-order structure, with CM and LM exhibiting a mutually antagonistic relationship, and LM exhibiting little interaction with OM. These results suggest distinct mechanisms for detecting edges defined by differences in orientation or by changes in contrast.

Acknowledgements: C.D. is funded by NIH grant R15-EY032732-01, C.B. is funded by Canadian NSERC grant RGPIN-2023-03559