Search

The registration of a 3D model over an image can be seen as the alignment of visual correspondences extracted from these two data. This is a challenging task and it is even more complex when the two images have a different modality. This paper introduces an approach that allows matching features detected in two different modalities: photographs and 3D models, by using a common 2D representation. Our approach is based on a modification of the Marching Cubes algorithm aiming to remove ambiguous cases without adding further calculations in each cube. We share the idea about the crucial importance of splitting the equivalence cases into two classes. Considering all the possible states inside/outside in the four corners of a cube side, indeed, there are only four non-trivial cases after eliminating those equivalences through the rotation. The obtained results allow us to validate the feasibility of the proposed methodology. [ABSTRACT FROM AUTHOR]