Not to be contrary, but I like the original (#3 above). The processed version (#2 above) looks over-cooked to me and makes me wonder if a lot of digital data got thrown out along the way. It might be interesting to watch the histogram during each step and track data loss through the process.