For computing a plurality of white balance gains with regard to a single image data item, a target image is first divided into a plurality of blocks. Then, a representative value of each block is computed, and a value representing the relative contribution of each light source type to a whole image is calculated based on the representative value which is obtained. Once the representative value of each block and the contribution level of each type of light source are obtained, a MWB gain and an AWB gain are computed using these results. Because the computation of the representative value of each block, which is a calculation operation performed for each pixel, is performed only once, the computation time can be greatly reduced.