Here are we mention several terms:
LiveWire - name of the tool for interactive delineation, the whole topic in those slides is about LiveWire if you wish. Dynamic Programming - the very common approach to solve optimization problems when the data has some sequential structure, as in our case.
Dijkstra algorithm -one of the most efficient methods to find the shortest path in the graph.
While in lecture it is explained in top-down strategy, from very common to practical, let me explain from the end.
We want to solve Live Wire problem: user gives two pixels, we need to find the "best" path. We define what is best for us: it is the path along the pixels with the highest gradient (it represents the contour). So we reduced our practical Live Wire problem to a shortest-path-search problem, which can be solved efficiently with Dijkstra. In turn, Dijkstra algorithm's core is dynamic programming: finding the best path using additional pixel using the best path without this pixel.