Symmetry-aware Registration of Human Faces

Prantl, Martin; Váša, Libor; Kolingerová, Ivana

doi:10.5220/0007254801850192

Cited by 2 publications

(1 citation statement)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As a consequence, we can only expect an exact match between two points in a symmetric pair in the case of synthetic or rounded data. Instead of perfect (exact, strong) symmetry, we thus consider weak (or approximate) symmetry [21,22]. This can be described formally by Equation (2).…”

Section: Introductionmentioning

confidence: 99%

AHiLS—An Algorithm for Establishing Hierarchy among Detected Weak Local Reflection Symmetries in Raster Images

Podgorelec,

Kolingerová,

Lovenjak

et al. 2024

Symmetry

View full text Add to dashboard Cite

A new algorithm is presented for detecting the local weak reflection symmetries in raster images. It uses contours extracted from the segmented image. A convex hull is constructed on the contours, and so-called anchor points are placed on it. The bundles of symmetry line candidates are placed in these points. Each line splits the plane into two open half-planes and arranges the contours into three sets: the first contains the contours pierced by the considered line, while the second and the third include the contours located in one or the other half-plane. The contours are then checked for the reflection symmetry. This means looking for self-symmetries in the first set, and symmetric pairs with one contour in the second set and one contour in the third set. The line which is evaluated as the best symmetry line is selected. After that, the symmetric contours are removed from sets two and three. The remaining contours are then checked again for symmetry. A multi-branch tree representing the hierarchy of the detected local symmetries is the result of the algorithm.

show abstract

Section: Introductionmentioning

confidence: 99%