An image processing device includes a recording control unit that controls recording of an image of a predetermined frame among images of a plurality of consecutive frames, a trim box setting unit that sets a plurality of trim boxes containing a subject area which is an area of a subject of interest in the image of the predetermined frame recorded by the recording control unit, and a generation control unit that controls generation of the plurality of trim images from the image of the predetermined frame recorded by the recording control unit, based on the plurality of trim boxes set by the trim box setting unit.