A specific target included in an image is detected by a specific target detection unit based on a specific target evaluation value representing likeliness of the specific target. Using a first white balance correction value calculation unit for obtaining a first white balance correction value for the region of the image other than the specific target, a second white balance correction value calculation unit for obtaining a second white balance correction value for the specific target region, and a white balance correction value comparison unit for making a comparison between the first and second white balance correction values, white balance in the region other than the specific target is corrected using the first white balance correction value, and white balance in the specific target region is corrected using the second white balance correction value, if the white balance correction values are different from each other.