An image processing device of the present invention has an area detection unit for detecting candidate areas for which structures of predetermined shapes are estimated to exist within an image acquired by image-capturing biological tissue a boundary pixel detection unit for detecting boundaries of the candidate areas an area setting unit for setting at least one local area upon each of both sides of a boundary within the vicinity of same a feature value calculation unit for calculating predetermined feature values on the basis of a predetermined value obtained for each pixel unit of each of the local areas that have been set by the area setting unit a discriminant value calculation unit for calculating a discriminant value on the basis of the calculation result of the predetermined feature value for the local area group of one side as seen from the boundary, and the calculation result of the predetermined feature value for the local area group of the other side as seen from the boundary and a candidate area correction unit for correcting the detection results of the candidate areas on the basis of the calculation results of the discriminant values.