An image processing device includes a facial region extraction unit extracting a facial region, an identification information acquisition unit acquiring identification information for identifying a face in the facial region, and first and second integrated processing units performing integrated processing. The first and second integrated processing units determine a threshold value on the basis of a relationship between an estimated area and a position of the face being tracked, calculate a similarity between a face being tracked and a face pictured in an image to be stored in a predetermined storage period, and determine if the face being tracked and the stored face image are the face of the same person.