A geometric approach to image labeling

• Verfasst von: Åström, Freddie [VerfasserIn]
• Verfasst von: Petra, Stefania [VerfasserIn]
• Verfasst von: Schmitzer, Bernhard [VerfasserIn]
• Verfasst von: Schnörr, Christoph [VerfasserIn]
• Titel: A geometric approach to image labeling
• Verf.angabe: Freddie Åström, Stefania Petra, Bernhard Schmitzer, Christoph Schnörr
• Umfang: 16 S.
• Fussnoten: Gesehen am 13.03.2018
• Titel Quelle: Enthalten in: Computer Vision – ECCV 2016
• Jahr Quelle: 2016
• Band/Heft Quelle: (2016), S. 139-154
• ISBN Quelle: 978-3-319-46454-1
• DOI: doi:10.1007/978-3-319-46454-1_9
• Datenträger: Online-Ressource
• Sprache: eng
• K10plus-PPN: 1570992339

Abstract

We introduce a smooth non-convex approach in a novel geometric framework which complements established convex and non-convex approaches to image labeling. The major underlying concept is a smooth manifold of probabilistic assignments of a prespecified set of prior data (the “labels”) to given image data. The Riemannian gradient flow with respect to a corresponding objective function evolves on the manifold and terminates, for any δ>0δ>0\delta > 0, within a δδ\delta -neighborhood of an unique assignment (labeling). As a consequence, unlike with convex outer relaxation approaches to (non-submodular) image labeling problems, no post-processing step is needed for the rounding of fractional solutions. Our approach is numerically implemented with sparse, highly-parallel interior-point updates that efficiently converge, largely independent from the number of labels. Experiments with noisy labeling and inpainting problems demonstrate competitive performance.