In an image processing device that discriminates between a mucosal region and a non-mucosal region that are contained in an intraluminal image, the image processing device includes a residue candidate region detecting unit that detects a residue candidate region to be discriminated whether the candidate region is a mucosal region or not, on the basis of a feature data of each pixel that constitutes the intraluminal image, a structure edge region detecting unit that detects a structure edge contained in the intraluminal image, and an approximate structure edge line calculating unit and an overlap deciding unit that discriminate whether the residue candidate region is a mucosal region or not, based on a relative positional relationship between the structure edge and the residue candidate region.