A Convex Geodesic Selective Model for Image Segmentation

Roberts, Michael ORCID: 0000-0002-3484-5031, Chen, Ke ORCID: 0000-0002-6093-6623 and Irion, Klaus L
(2019) A Convex Geodesic Selective Model for Image Segmentation. Journal of Mathematical Imaging and Vision, 61 (4). 482 - 503.

[img] Text
MikeP2_Draft.pdf - Accepted Version

Download (13MB)


Selective segmentation is an important application of image processing. In contrast to global segmentation in which all objectsare segmented, selective segmentation is used to isolate specific objects in an image and is of particular interest in medicalimaging—permitting segmentation and review of a single organ. An important consideration is to minimise the amount of userinput to obtain the segmentation; this differs from interactive segmentation in which more user input is allowed than selectivesegmentation. To achieve selection, we propose a selective segmentation model which uses the edge-weighted geodesicdistance from a marker set as a penalty term. It is demonstrated that this edge-weighted geodesic penalty term improveson previous selective penalty terms. A convex formulation of the model is also presented, allowing arbitrary initialisation.It is shown that the proposed model is less parameter dependent and requires less user input than previous models. Furthermodifications are made to the edge-weighted geodesic distance term to ensure segmentation robustness to noise and blur. Wecan show that the overall Euler–Lagrange equation admits a unique viscosity solution. Numerical results show that the resultis robust to user input and permits selective segmentations that are not possible with other models.

Item Type: Article
Uncontrolled Keywords: Variational model, Partial differential equations, Image segmentation, Additive operator splitting, Viscosity solution, Geodesic
Depositing User: Symplectic Admin
Date Deposited: 18 Oct 2018 09:49
Last Modified: 29 Oct 2021 14:10
DOI: 10.1007/s10851-018-0857-2
Open Access URL: https://link.springer.com/article/10.1007/s10851-0...
Related URLs:
URI: https://livrepository.liverpool.ac.uk/id/eprint/3027694