The algorithm is relatively simple. The points are sorted by their polar angle relative to the reference point, keeping track of which line segment they belong to. In this ordering, they are assigned tags representing whether they are a start of a segment, or it's end. A comparison is defined on those oriented segments that determines which segment covers the other if viewed from the reference point. Then the list of points is folded over by putting segments in a set, and removing them from it when a starting point or an end point is encountered, respectively. Every time the closest segment to the reference point changes, we output two points of a visibility polygon by computing relevant ray-segment intersections.