Impossible Figures, such as the Penrose triangle or those created by M.C.
Escher, are a form of optical illusion consisting of locally possible sections
joined together to form a globally inconsistent structure. Through exploiting
hidden deformations within the object’s structure it is possible to model
three-dimensional representations of these impossible figures. The hidden
deformation methods used, however, result in corresponding deformations
or anomalies in the object’s cast shadow. Due to these deformed shape
of the resulting shadows they are often excluded from the final rendering.
By excluding cast shadows entirely the important contextual and positional
information they contain about the scene is also lost.
This work presents a novel method of modeling three-dimensional versions
of these impossible figures through the use of transparency. By manipulating
the surface transparency we demonstrate the ability to simulate a
range of impossible figures under new viewpoints. We also produce a screen
space occlusion method for rendering impossible cast shadows automatically.
By manipulating information stored within the depth buffer our algorithm
produces impossible cast shadows for both possible and impossible objects.
Providing a solution to casting visually appropriate shadows for impossible
objects, along with other rendering features such as ambient occlusion. By
operating in screen space our algorithm works without disrupting the existing
rendering pipeline, as such we have implemented it within the Unity engine.
To examine the effectiveness of our copycat shadow algorithm we conduct
of visual perception experiments. Exploring the effect cast shadows have on
the viewers perception of impossible objects.