The marching squares algorithm works, but it doesn't 'find' any holes or gaps that might exist in a discovered blob of collision tiles. This is a perfectly viable scenario after all, you might have pockets of navigable areas in your map (eg, a ring of collision tiles with a square of viable terrain in the middle).
Ideas?
Maybe reverse the calculations and instead of finding blobs of colliding tiles, find the blobs of non-colliding areas: that's what we want after all to put into the delaunay afterwards anyway. May need to add a phantom border around the edge of the map of colliding tiles so the calculation can 'find' all the way out to the edge.
Problem
The marching squares algorithm works, but it doesn't 'find' any holes or gaps that might exist in a discovered blob of collision tiles. This is a perfectly viable scenario after all, you might have pockets of navigable areas in your map (eg, a ring of collision tiles with a square of viable terrain in the middle).
Ideas?