Introduction to the Random Walker Algorithm: Graph-Theoretic Electrical Potentials for Multi-Label Medical Image Segmentation

The random walker algorithm, initially proposed by Leo Grady and Gareth Funka-Lea, is a graph-based segmentation technique that models image pixels as nodes in a graph and computes electrical potentials to achieve multi-label segmentation. This method formulates segmentation as a combinatorial Dirichlet problem solvable through sparse linear systems, making it particularly effective for medical imaging applications with weak boundaries and noise.