Combining Geometric and Topological Information for Boundary Estimation
Hengrui Luo, Justin Strait · 2021 IEEE International Conference on Big Data (Big Data) · 2021
We propose a method which jointly incorporates geometric and topological information to simultaneously estimate boundaries for objects in images with more complex topologies. We use a topological clustering-based method to assist the initialization of the Bayesian active contour model. When applied separately, the topological clustering is not robust to background noise, while active contour methods are known to be extremely sensitive to algorithm initialization. Our proposed topologically guided method provides an interpretable, principled initialization in these settings, which avoids potential pitfalls associated with these types of objects. We provide a simulation study comparing our initialization to boundary estimates obtained from standard segmentation algorithms on simulated images, and then demonstrate our method successfully on real-world applications with skin lesions and neural cellular images, for which multiple topological features can be identified automatically.