A three dimensional image processing apparatus includes a region-of-interest setting unit to obtain a moving image of an area of movement of the subject radiographed from a predetermined direction, the moving image including a sequence of frames, to obtain a first feature region in a first frame of the sequence, to search each subsequent frame in the sequence to identify corresponding feature regions most similar to the first feature region so as to identify a sequence of feature regions, and to set a region of interest on the first frame as a rectangular region that just covers all of the identified feature regions in the sequence of feature regions.