Fast and exact visibility on digitized shapes and application to saliency-aware normal estimation
This work addresses computational efficiency in geometric processing for digital shapes, with applications in computer graphics and vision, but it is incremental as it builds on existing digital set representations.
The authors tackled the problem of efficiently computing visibility graphs on digital shapes by using a representation based on lists of integral intervals, enabling fast and exact visibility determination. As an application, they demonstrated accurate and convergent normal vector field estimation that preserves sharp features of the shape.
Computing visibility on a geometric object requires heavy computations since it requires to identify pairs of points that are visible to each other, i.e. there is a straight segment joining them that stays in the close vicinity of the object boundary. We propose to exploit a specic representation of digital sets based on lists of integral intervals in order to compute eciently the complete visibility graph between lattice points of the digital shape. As a quite direct application, we show then how we can use visibility to estimate the normal vector eld of a digital shape in an accurate and convergent manner while staying aware of the salient and sharp features of the shape.