The section on shape mentions that non-isotropic textures result in ambiguous slant measurements. However, this does not seem to be a problem in practice for perception, even though many real-world textures are not isotropic (such as the tree bark or the grass in Fig 8.2). What accounts for our perceptual resiliency to these ambiguities? Is it prior knowledge of the textures in question? Some intrinsic aspect of the textures? Or do we factor in other shape cues?
Maybe we're just comfortable with statistical local measures in images? I am surprised the evidence is inconclusive about our assumption of isotropy in textures -- for example, the appearance of shape in the first image in the chapter suggests that a distorted pink noise texture is sufficient to give the illusion of shape. This area seems substantially under-explored.
The section on shape mentions that non-isotropic textures result in ambiguous slant measurements. However, this does not seem to be a problem in practice for perception, even though many real-world textures are not isotropic (such as the tree bark or the grass in Fig 8.2). What accounts for our perceptual resiliency to these ambiguities? Is it prior knowledge of the textures in question? Some intrinsic aspect of the textures? Or do we factor in other shape cues?
ReplyDeleteMaybe we're just comfortable with statistical local measures in images? I am surprised the evidence is inconclusive about our assumption of isotropy in textures -- for example, the appearance of shape in the first image in the chapter suggests that a distorted pink noise texture is sufficient to give the illusion of shape. This area seems substantially under-explored.
ReplyDelete