A medical image processing device includes an input section to which a biological mucous membrane image obtained by picking up an image of a biological mucous membrane is inputted, a region extracting section that extracts a mucous membrane microstructure region corresponding to a mucous membrane microstructure from the inputted biological mucous membrane image, a closed region identifying section that identifies at least one closed region regarded as being surrounded by the mucous membrane microstructure region, and a unit region setting section that sets a biologically histological unit region on the basis of the mucous membrane microstructure region and the closed region.