An image generation device includes: a detection unit 43 that performs processing of detecting a specific feature from a plurality of focus images which include an observation target and are in different focus states; a determination unit 44 that determines a parameter indicating a degree of application of a region image to a combination region image in a case where the combination region image is generated from the region image, for each set of the region images in a plurality of corresponding regions respectively corresponding to the plurality of focus images, based on the specific feature detected by the detection unit 43; and a generation unit 45 that generates the combination region image by combining the region images for each of the plurality of corresponding regions based on the parameter determined by the determination unit 44.