An image processing apparatus includes a feature-quantity calculating unit that calculates feature quantities of target regions each indicating a tracking object in respective target images, the target images being obtained by capturing the tracking object at a plurality of time points; a provisional-tracking processing unit that performs provisional tracking of the target region by associating the target regions of the target images with each other using the calculated feature quantities; and a final-tracking processing unit that acquires a final tracking result of the target region based on a result of the provisional tracking.