[Problem] To extract, with high precision, regions of residue in a hollow organ from a luminal image that includes a hollow organ such as the large intestine. [Solution] A region extraction device comprises: a gaseous region extraction unit (11) which extracts gaseous regions from a luminal image that includes a hollow organ a residue candidate region extraction unit (12) which extracts candidates of regions of residue in the hollow organ from the luminal image as residue candidate regions a boundary candidate region detection unit (13) which detects boundary candidate regions that include boundaries between the gaseous regions and the residue candidate regions a representative direction component acquisition unit (14) which acquires a representative direction component that represents multiple direction components in the image of the boundary candidate regions a boundary region detection unit (15) which, on the basis of the representative direction component, detects boundary regions that include the boundaries between the gaseous regions and the regions of residue from the boundary candidate regions and a residue region extraction unit (16) which extracts, as regions of residue, the residue candidate regions that include the boundary regions.La présente invention vise à extraire, avec une haute précision, des régions de résidu dans un organe creux à partir dune image luminale qui comprend un organe creux, tel que le gros intestin. A cet effet, linvention concerne un dispositif dextraction de région qui comprend : une unité dextraction de région gazeuse (11) qui extrait des régions gazeuses dune image luminale qui comprend un organe creux une unité dextraction de région candidate de résidu (12) qui extrait, en tant que régions candidates de résidu, des régions candidates de résidu dans lorgane creux à partir de limage luminale une unité de détection de région candidate de limite (13) qui détecte des régions candidates de limite qui comprennent des limites entre les
