An image processing apparatus includes at least one processor configured to execute processes including: calculating a misalignment amount of each pixel of a reference image relative to a standard image; and combining the reference image converted based on the calculated amount with the standard image. The calculating includes: calculating a projection conversion matrix for each of planes with different misalignment amounts in the reference image; generating a plane map in which the plane to which each pixel of the reference image belongs and the matrix to be applied to each plane are selected based on a difference value between the standard image and each of the alignment images converted from the reference image by using each calculated matrix; suppressing a selection error of the matrix; and calculating the misalignment amount for each of the planes based on the plane map in which the selection error of the matrix is suppressed.