In order to detect retention in a preferable manner, an image processing system is provided with: a retention area extraction unit that determines whether an area is a retention area in an image frame of a processing time on the basis of a first image generated from each of image frames taken within a first time width from the processing time and a second image generated from each of image frames taken within a second time width from the processing time which is longer than the first time width; and a reliability calculation unit that generates reliability information relating to the determination of the retention area for each area in the image frames to be processed.