Provided is an image processing device, and the like, that can prevent abnormal images in which the same abnormal region is shown from being consecutively extracted as a representative image, in a case where a representative image is extracted from a series of images acquired by picking up images in time-sequence order of the inside of a lumen of an organism. An image processing device (1) comprises: a detection unit (110) that detects abnormal images, which include an abnormal region, from a series of images acquired by sequentially picking up images of the inside of the lumen of the organism; an overall similarity calculation unit (120) that, among different abnormal-images, calculates a general similarity which is the similarity among regions which include regions other than abnormal regions; an abnormal image group extraction unit (130) that extracts an abnormal image group the images of which include the same abnormal region, on the basis of a comparison between a threshold value and the general similarity or a discrimination parameter based on such general similarity; and a representative image extraction unit (140) that extracts a representative image from the extracted abnormal image group.L'invention concerne un dispositif de traitement d'image, et analogue, qui peut empêcher des images anormales, dans lesquelles la même région anormale est illustrée, d'être extraites de façon consécutive sous la forme d'une image représentative, dans un cas où une image représentative est extraite d'une série d'images acquises par capture d'images dans un ordre de séquence temporelle de l'intérieur d'une lumière d'un organisme. Un dispositif de traitement d'image (1) comprend : une unité de détection (110) qui détecte des images anormales, qui comprennent une région anormale, à partir d'une série d'images acquises par capture séquentielle d'images de l'intérieur de la lumière de l'organisme ; une unité de calcul de similarité globale (120) qui, parmi différentes images ano