A robust and rotationally invariant local surface descriptor with applications to non-local mesh processing

In recent years, we have witnessed a striking increase in research concerning how to describe a meshed surface. These descriptors are commonly used to encode mesh properties or guide mesh processing, not to augment existing computations by replication. In this work, we first define a robust surface descriptor based on a local height field representation, and present a transformation via the extraction of Zernike moments. Unlike previous work, our local surface descriptor is innately rotationally invariant. Second, equipped with this novel descriptor, we present SAMPLE – similarity augmented mesh processing using local exemplars – a method which uses feature neighbourhoods to propagate mesh processing done in one part of the mesh, the local exemplar, to many others. Finally, we show that SAMPLE can be used in a number of applications, such as detail transfer and parameterization.

Sample: The parameterization done on the selected dragon scale (bottom right) is propagated to several other similar dragon scales,the same texture (bottom left) is mapped to each region.Figure optionsDownload as PowerPoint slideHighlights
► A robust surface descriptor based on local heightmaps.
► Zernike moments transformation to make the descriptor rotationally invariant.
► SAMPLE – similarity augmented mesh processing using local exemplars.
► SAMPLE is used to propagate mesh processing throughout similar regions.