A Single Directrix Quasi-Minimal Model for Paper-Like Surfaces

We are interested in reconstructing paper-like objects from images. These objects are modeled by developable surfaces and are mathematically wellunderstood. They are difficult to minimally parameterize since the number of meaningful parameters is intrinsically dependent on the actual surface. We propose a quasi-minimal model which self-adapts its set of parameters to the actual surface. More precisly, a varying number of rules is used jointly with smoothness constraints to bend a flat mesh, generating the sought-after surface. We propose an algorithm for fitting this model to multiple images by minimizing the point-based reprojection error. Experimental results are reported, showing that our model fits real images accurately.