A Novel Riemannian Framework for Shape Analysis of Annotated Surfaces

Jiaqi Zaetz and Sebastian Kurtek

Abstract

We present a novel, parameterization-invariant method for shape analysis of annotated surfaces. While the method can handle various types of annotation including color and texture, in this paper we focus on soft landmark annotations. Landmark annotations are commonly provided in various applications including medical imaging where an expert marks points of interest on the objects. Most methods in current literature either study shapes using landmarks only or surfaces only. In either case, the analyst is forced to ignore a lot of useful information. We propose a novel representation of surfaces that can jointly incorporate shape and landmark annotation. Our framework properly removes all shape preserving transformations from the representation space including translation, scale, rotation and re-parameterization. We present results of comparing and averaging annotated shapes for toy data as well as handwritten digits represented as graph surfaces. We achieve good classification results on the handwritten digits as well as Attention Deficit Hyperactivity Disorder (ADHD) using subcortical structures.

Session

Workshop: 1st International Workshop on DIFFerential Geometry in Computer Vision for Analysis of Shapes, Images and Trajectories (DIFF-CV)

Files

PDF iconPaper (PDF, 5M)

DOI

10.5244/C.29.DIFFCV.3
https://dx.doi.org/10.5244/C.29.DIFFCV.3

Citation

Jiaqi Zaetz and Sebastian Kurtek. A Novel Riemannian Framework for Shape Analysis of Annotated Surfaces. In H. Drira, S. Kurtek, and P. Turaga, editors, Proceedings of the 1st International Workshop on DIFFerential Geometry in Computer Vision for Analysis of Shapes, Images and Trajectories (DIFF-CV 2015), pages 3.1-3.11. BMVA Press, September 2015.

Bibtex

@inproceedings{DIFFCV2015_3,
	title={A Novel Riemannian Framework for Shape Analysis of Annotated Surfaces},
	author={Jiaqi Zaetz and Sebastian Kurtek},
	year={2015},
	month={September},
	pages={3.1-3.11},
	articleno={3},
	numpages={11},
	booktitle={Proceedings of the 1st International Workshop on DIFFerential Geometry in Computer Vision for Analysis of Shapes, Images and Trajectories (DIFF-CV 2015)},
	publisher={BMVA Press},
	editor={H. Drira, S. Kurtek, and P. Turaga},
	doi={10.5244/C.29.DIFFCV.3},
	isbn={1-901725-56-1},
	url={https://dx.doi.org/10.5244/C.29.DIFFCV.3}
}