by Jiri Hladuvka, Florian Bogner, Walter G. Kropatsch
Abstract:
The practical use of generalized combinatorial maps(n-Gmaps) to represent nD-images is limited by large memory requirements. Implicit representations, if well designed, can come to the rescue. Unbounded n-Gmaps are surrounded by an infinite background region that renders pixel-oriented implicit representation schemes inconvenient. The contribution of this paper is twofold. First, we propose an implicit, membrane-centric (rather than pixel-centric) arrangement of darts in unbounded n-Gmaps. Second, we introduce involutions based on bit flips, which allow efficient iterations in membranes and have the potential to speed up computations. We have validated our approach on a variety of 2D and 3D images, including those whose memory requirements would far exceed the main memory.
Reference:
Implicit unbounded n-Gmaps for images: A membrane-centric encoding using bit-flips (Jiri Hladuvka, Florian Bogner, Walter G. Kropatsch), Technical report, PRIP, TU Wien, 2022.
Bibtex Entry:
@TechReport{TR154,
author = "Jiri Hladuvka and Florian Bogner and Walter G. Kropatsch",
title = "Implicit unbounded n-Gmaps for images: A membrane-centric encoding using bit-flips",
institution = "PRIP, TU Wien",
number = "PRIP-TR-154",
year = "2022",
url = "https://www.prip.tuwien.ac.at/pripfiles/trs/tr154.pdf",
abstract = "The practical use of generalized combinatorial maps(n-Gmaps) to represent nD-images is limited by large memory requirements. Implicit representations, if well designed, can come to the rescue. Unbounded n-Gmaps are surrounded by an infinite background region that renders pixel-oriented implicit representation schemes inconvenient. The contribution of this paper is twofold. First, we propose an implicit, membrane-centric (rather than pixel-centric) arrangement of darts in unbounded n-Gmaps. Second, we introduce involutions based on bit flips, which allow efficient iterations in membranes and have the potential to speed up computations. We have validated our approach on a variety of 2D and 3D images, including those whose memory requirements would far exceed the main memory.",
}