For longer paths, you would need to connect the end of an eye path to the end of a light path, right? My first guess was that this involves having to store the endpoints of eye paths and the endpoints of light paths, but that seems really expensive in terms of memory. What sorts of techniques are used in practice to join longer paths?
bcagan
Does this end up being the most common solution to reducing variance? It seems less costly to than the others to combine these but it also doesn't seem like that would be as effective.
For longer paths, you would need to connect the end of an eye path to the end of a light path, right? My first guess was that this involves having to store the endpoints of eye paths and the endpoints of light paths, but that seems really expensive in terms of memory. What sorts of techniques are used in practice to join longer paths?
Does this end up being the most common solution to reducing variance? It seems less costly to than the others to combine these but it also doesn't seem like that would be as effective.