Abstract
Recent thrust in imaging capabilities in medical as well as emerging areas ofmanufacturing systems creates unique opportunities and challenges foron-the-fly, unsupervised estimation of anomalies and other regions of interest.With the ever-growing image database, it is remarkably costly to createannotations and atlases associated with different combinations of imagingcapabilities and regions of interest. To address this issue, we present anunsupervised learning approach to a continuous max-flow problem. We show thatthe maximum a posteriori estimation of the image labels can be formulated as acapacitated max-flow problem over a continuous domain with unknown flowcapacities. The flow capacities are then iteratively obtained by considering aMarkov random field prior over the neighborhood structure in the image. We alsopresent results to establish the consistency of the proposed approach. Weestablish the performance of our approach on two real-world datasets including,brain tumor segmentation and defect identification in additively manufacturedsurfaces as gathered from electron microscopic images. We also present anexhaustive comparison with other state-of-the-art supervised as well asunsupervised algorithms. Results suggest that the method is able to performalmost comparable to other supervised approaches, but more 90% improvement interms of Dice score as compared to other unsupervised methods.