An image processing apparatus of the invention includes: a region detection section which detects a candidate region in which a structure having a predetermined shape is estimated to exist from an image obtained by image-pickup of a living tissue a border pixel detection section which detects a border of the candidate region a region setting section which sets at least one local region in the vicinity of and both sides of the border a feature value calculation section which calculates predetermined feature values based on predetermined values obtained for respective pixel units of the local regions set by the region setting section a discrimination value calculation section which calculates a discrimination value based on a calculation result of the predetermined feature values in a local region group on one side when viewed from the border and a calculation result of the predetermined feature values in a local region group on the other side when viewed from the border and a candidate region correction section which corrects a detection result of a candidate region based on a calculation result of the discrimination value.
