In an image acquisition device, when capturing an optical image of a sample S through lane scanning, the number of tile images T included in a tile image row R acquired in one lane is counted, and a determination is made as to whether or not the number of images reaches a planned acquisition count that is set in advance. In a case where it is determined that the number of images does not reach the planned acquisition count, lane scanning with respect to the lane is re-executed. Accordingly, even when a loss of the tile images T occurs due to an environment load, the tile images T are complemented by re-execution of the lane scanning, and thus it is possible to prevent the loss of the tile images T in an observation image.