Abstract
Selective segmentation is an important application of image processing. In contrast to global segmentation in which all objects are segmented, selective segmentation is used to isolate specific objects in an image and is of particular interest in medical imaging—permitting segmentation and review of a single organ. An important consideration is to minimise the amount of user input to obtain the segmentation; this differs from interactive segmentation in which more user input is allowed than selective segmentation. To achieve selection, we propose a selective segmentation model which uses the edge-weighted geodesic distance from a marker set as a penalty term. It is demonstrated that this edge-weighted geodesic penalty term improves on 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. Further modifications are made to the edge-weighted geodesic distance term to ensure segmentation robustness to noise and blur. We can show that the overall Euler–Lagrange equation admits a unique viscosity solution. Numerical results show that the result is robust to user input and permits selective segmentations that are not possible with other models.
Highlights
Segmentation of an image into its individual objects is one incredibly important application of image processing techniques
Their model uses Euclidean distance from the marker set M as a distance penalty term; we propose replacing this with the edge-weighted geodesic distance from M
We propose a convex selective segmentation model using this penalty term and demonstrate how it can achieve results which cannot be achieved by other models
Summary
Segmentation of an image into its individual objects is one incredibly important application of image processing techniques. We refer to [5,23] for selective segmentation models which include different fitting constraints, using coefficient of variation and the centroid of M, respectively None of these models have a restriction on the size of the object or objects to be detected, and depending on the initialisation, these methods have the potential to detect more or fewer objects than the user desired. Spencer and Chen [38], in the same paper, reformulated the model they introduced to a convex form using convex relaxation and an exact penalty term as in [10] Their model uses Euclidean distance from the marker set M as a distance penalty term; we propose replacing this with the edge-weighted geodesic distance from M (we call this the geodesic distance).
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have