An image reading apparatus for reading by an image sensing element an original held by an original holder while performing pixel shifting operation detects the feature point of the original holder, specifies the original position on the basis of the detected feature point, and performs the first reading of the original. After reading, the apparatus shifts the relative positions of the optical image of the original and the image sensing element by a predetermined amount. Then, the apparatus detects the feature point of the original holder again, specifies the original position on the basis of the detected feature point, and performs the second reading of the original. The apparatus synthesizes the two read original images to generate one image.