It is a proposition to reduce user's burden when performing tracking observation of a specimen. Therefore, an image processing apparatus includes a storage unit which takes in a plurality of slice images acquired from a specific area of a specimen by each of Z-stack shootings performed at different time point, associates the plurality of slice images each with a shooting time point and slice height z, and stores the plurality of slice images as one of slice image groups, a manual selecting unit which allows user to manually select a representing image (Ia, Ib, Ic, . . . ) from one of slice image groups, and an automatic selecting unit which automatically selects a representing image from another one of slice image groups based on a result of the manual selection.