I assumed the OP was talking more about the math, which is also not true. The method mentioned in the article only takes into account the silhouette of the targ

I assumed the OP was talking more about the math, which is also not true. The method mentioned in the article only takes into account the silhouette of the target. In my experience, that's where a lot of the information comes from, but to get that fine detail most contemporary methods incorporate a lot of information about the target itself like keypoints and lighting interactions.