An image matching device in which, in automatic positioning calculation using an ICP method, a solution is converged to an optimal solution while avoiding reaching a local solution, and a patient positioning device using this matching device. The device includes a CT image data reading unit reading first and second CT image data, a point group data generation unit generating first and second point group data in a three-dimensional space by binarization processing and edge-extraction processing on sliced images of the first and second CT image data, respectively, a point group data resolution conversion unit thinning-out point group data so as to extend a point-group-data array pitch, and an ICP calculation unit obtaining, using an ICP method, a conversion amount for the second point group data so that an error function of the first point group data and the second point group data outputted from the point group data resolution conversion unit is minimized.