On 2016 May 29, Lydia Maniatis commented:

Here's the thing. First, describing an image as a case of "shape-from-shading" is jumping the gun. Pre-viewing the image contains luminance variations which may or may not be interpreted as changes in illumination. This should be obvious when we are using pictorial rather than real-world stimuli. So the question is when are the changes in luminance seen as changes in illumination? The answer is that we (our visual system, in effect) judge the possible shapes that will arise under various reflectance/illumination options

If we take a clear "shape-from-shading" figure and we make the edges of the shadows hard, make the shadowed areas solid black, or remove the luminance changes but simply trace them out with a hard line, we will in most cases still see the same 3D shapes; it will be a kind of cartoon version of the shape-from-shading impression. We'll have in effect contour lines. This will happen b/c treating the lines in any other way (as delineating shapes in themselves) will produce worse shapes, specifically shapes will an implicitly smaller volume/area ratio. So "shape-from-shading" should be called "shading-from-shape."

On 2016 May 25, Lydia Maniatis commented:

When Todd et al say that understanding their demonstrations requires “a broader theoretical analysis of shape from shading that is more firmly grounded in ecological optics,” do they mean that there are things about the physics of how light interacts with surfaces that we don't understand? What kind of empirical investigations are they suggesting need to be performed? What kind of information do they think is missing, optically-speaking?

The fundamental issues are formal (having to do with form) not the details of optics and probabilities of illumination structure - as these authors have shown.

On 2016 May 25, Lydia Maniatis commented:

When Todd et al say that understanding their demonstrations requires “a broader theoretical analysis of shape from shading that is more firmly grounded in ecological optics,” do they mean that there are things about the physics of how light interacts with surfaces that we don't understand? What kind of empirical investigations are they suggesting need to be performed? What kind of information do they think is missing, optically-speaking?

The fundamental issues are formal (having to do with form) not the details of optics and probabilities of illumination structure - as these authors have shown.

On 2016 May 29, Lydia Maniatis commented:

Here's the thing. First, describing an image as a case of "shape-from-shading" is jumping the gun. Pre-viewing the image contains luminance variations which may or may not be interpreted as changes in illumination. This should be obvious when we are using pictorial rather than real-world stimuli. So the question is when are the changes in luminance seen as changes in illumination? The answer is that we (our visual system, in effect) judge the possible shapes that will arise under various reflectance/illumination options

If we take a clear "shape-from-shading" figure and we make the edges of the shadows hard, make the shadowed areas solid black, or remove the luminance changes but simply trace them out with a hard line, we will in most cases still see the same 3D shapes; it will be a kind of cartoon version of the shape-from-shading impression. We'll have in effect contour lines. This will happen b/c treating the lines in any other way (as delineating shapes in themselves) will produce worse shapes, specifically shapes will an implicitly smaller volume/area ratio. So "shape-from-shading" should be called "shading-from-shape."