Abstract
How do we effortlessly untangle complex mashes of image structure to visually infer the shape and surface quality of 3D objects? We explored whether different surface optics improve global shape perception. We created virtual objects varying in surface optics (matte, specular, refractive (with/without specular reflection)), relief (smooth, bumpy), and stretch in shape along the viewing axis. Rendering were performed in the natural environments (Eucalyptus Grove, and St. Peter's Basilica, Debevec 1998). The objects oscillated in 1 Hz horizontally and displayed for 5 s on a flat screen. Eight observers used an analog scale to match the shape of the 3D object's appearance as elongated (like a rugby ball) or flat (like a pancake). No response feedback was given. We found that the flat objects with specular reflection were perceived elongated compared with matte surfaces (similar to Mooney and Anderson, 2014). Any surfaces containing specular components with or without refractive components showed similar results. We also found that the refractive objects without specular reflection were perceived flatter than the other objects for all elongated shapes. These effects were consistently observed for objects of different sizes, binocular and monocular viewing, and even for static images. Refractive index (RI) distorts the image structure necessary for 3D shape perception, so we expected perceived shape to vary with RI. However, we observed that 3D shape perception was significantly 'flatter' for refractive materials, compared with purely diffuse and specular materials. The observer underestimates of convexity in refractive objects can be explained by velocity of the optic flow field.
Meeting abstract presented at VSS 2018