An image processing device includes a motion-vector detecting unit that detects motion vectors between an image and another image at a plurality of positions on the image, the image being at least a part of a plurality of images making up an image sequence; a pattern categorizing unit that categorizes the images into pattern groups each indicating a predetermined motion pattern, on the basis of the plurality of the motion vectors detected in the image; a group-based image-change detecting unit that detects, in each of the pattern groups, an image change between the images in the pattern group; and a group-based scene-change detecting unit that detects scene change in each of the pattern groups on the basis of the image change between the images in the pattern group.