To reduce a processing time for computing a composition position without lowering superposition accuracy in image processing for forming a composite image from partial images.

When a partial image from a sensor 1 is stored in a current frame memory 1, a region dividing section 13 divides the partial image into a plurality of regions. A feature region retrieval section 14 computes the feature quantity (e.g. the number of black pixels continuing in a lateral direction) included in the regions for each divided region, and detects the region with the largest feature quantity as a feature region. A region pixel comparing section 15 reads image data of the specified feature region respectively from the current frame memory 11 and a previous frame memory 12, superposes them by shifting one pixel, computes the total difference of luminance for each pixel in the superposed part with respect to each superposed position, and outputs the superposed position where the total value is minimum, as the composition position to a superposing section 16.