Generative constraints on image cues for perceived gloss

摘要: Image structure is generated by distinct physical sources that include 3-D surface geometry, reflectance, and the light field. A fundamental problem in midlevel vision involves understanding how, how well, visual system separates images into their underlying sources. large body of recent work has shown perception gloss exhibits significant dependencies on both geometry field which embedded. We recently proposed these arise from relying a restricted set dimensions specular image are heuristically used to estimate (Marlow, Kim, & Anderson, 2012). Here, we report series studies manipulate prospectively test whether cues predict perceived for broad range geometries fields. identify generative constraints each our allow us cue will vary most given reflectance function, Our psychophysical data reveal manipulations succeed promoting significance intended cue, turn increases predictive value observers' judgments gloss. results suggest static, monocular relies heuristic weighting structure, offer unified explanation why modulate