Shape Segmentation for Mesh-Based 3D Models: Unifying Part- and Patch-type Approaches

Koehoorn, J (2016) Shape Segmentation for Mesh-Based 3D Models: Unifying Part- and Patch-type Approaches. Master's Thesis / Essay, Computing Science.

Preview	Text thesis.pdf - Published Version Download (32MB) | Preview
	Text Toestemming.pdf - Other Restricted to Backend only Download (78kB)

Abstract

Shape segmentation algorithms can be categorized into two main categories, namely part-based segmentation and patch-based segmentation. The former identifies the structural parts of a shape, much like a human would visually classify parts, and is most suitable for natural shapes, whereas patch-based segmentation concerns itself with partitioning a shape into its quasi-flat areas which best suits man-made objects. This dichotomy poses an issue with being able to automatically segment any input shape, as it is unknown which method to use. Our contribution is two-fold: we first present a new part-based segmentation method to operate on mesh shapes, derived from cut-space segmentation as proposed by Feng et al in 2015. Secondly, we propose a heuristic to automatically select what segmentation is most appropriate without introducing undesirable over-segmentation. To accomplish this we first validate each part of the part-based segmentation based on several structural properties a part must satisfy, then determine for each remaining parts what patches should be introduced. We tested our approach on over fifty various kinds of shapes and obtained promising results. We show some of the results and also discuss shapes that are problematic with our algorithm.

Item Type:	Thesis (Master's Thesis / Essay)
Degree programme:	Computing Science
Thesis type:	Master's Thesis / Essay
Language:	English
Date Deposited:	15 Feb 2018 08:26
Last Modified:	15 Feb 2018 08:26
URI:	https://fse.studenttheses.ub.rug.nl/id/eprint/14723

Actions (login required)

View Item