[Back]

D. Shao, W. Kropatsch:
"Irregular {L}aplacian {G}raph {P}yramid";
Talk: Computer Vision Winter Workshop 2010, Nove Hrady; 2010-02-03 - 2010-02-05; in: "Proceedings of the 15th Computer Vision Winter Workshop", Libor {\v S}pa{\v c}ek and Vojt{\v e}ch Franc, Czech Society for Cybernetics and Informatics, and the Center of Machine Perception at CTU Prague/Nov\'e Hrady, Cz (2010), ISBN: 978-80-254-6499-1; 66 - 71.

Abstract This paper presents a novel image representation,
which incorporates the principles of Laplacian Pyramid
into the irregular graph pyramid. The drawback of the
regular Laplacian Pyramid is their lack to keep the topological
structure of the image, due to the contraction process in
building the Gaussian Pyramid. Irregular graph pyramid is
able to hierarchically represent the topological structure of
an image with multiresolution, where each level is a graph
describing the image with various resolutions by contracting
the graph from the level below. We build irregular Laplacian
graph pyramid by storing the difference of levels in irregular
graph pyramid. Experiments and results are presented in the
paper to show the characteristic of the irregular Laplacian
graph pyramid and some immediate advantages in computer
vision applications.