A same subject of imaging is captured continuously, one frame at a time, from plural viewpoints by imaging units, and a subject is detected from images of the captured frames, and, if plural subjects are detected, a range represented by a difference between a maximum value and a minimum value of distances between the imaging units and the subjects is computed. If a difference between a range of a specific frame and a range of a frame one before or after exceeds a predetermined threshold value, the range of the specific frame is adjusted such that the difference becomes smaller. Based on a predetermined relationship between ranges and parallax amounts corresponding to the ranges, a parallax amount corresponding to the computed range or the adjusted range is computed, and the parallax amounts and the images of the respective frames are recorded in correspondence with one another on a recording unit.