Shape from darkness: deriving surface information from dynamic shadows

摘要: We present a new method, shape from darkness, for extracting surface information based on object self-shadowing under moving light sources. It is motivated by the problem of human perception fractal textures perspective. One-dimensional dynamic shadows are analyzed in continuous case, and their behavior categorized into three exhaustive shadow classes. The shown to be solved integration ordinary differential equations, using captured image representation called suntrace. discretization one-dimensional introduces uncertainty discrete suntrace; however it successfully recast as satisfaction 8n constraint equations 2n unknowns. A form relaxation appears quickly converge these constraints accurate reconstructions; we give several examples simulated images. darkness method has two advantages: does not require reflectance map, works non-smooth surfaces. conclude with discussion method's accuracy practicality, its relation perception, future extensions.