An image processing apparatus comprises a corresponding point determining unit configured to, for a plurality of points contained in a first image, search a second image for corresponding points; a transformation coefficient calculating unit configured to divide the plurality of corresponding points into groups, based on amounts of misalignment between the images at the corresponding points, and configured to calculates a coordinate transformation coefficient for each of the groups; and an image synthesizing unit configured to generate a synthesis image, using a plurality of the coordinate transformation coefficients and the second image.