A new morphological image processing algorithm is proposed for automatically and non-destructively analyzing the cellular structure of planar-type bioceramic foam using an optical microscope. With an abundant cell density, the cell structure can differentiate high cell volume fraction, high degree of interconnection to neighboring cells, and variable cell sizes. Our primary focus was to identify whether the distribution in cell sizes could be reliably measured by automatically delineating cell boundaries on the basis of a difference in the optical reflection of the cell and strut. There should be no mechanical or chemical pretreatment needed to enhance their visibility. Under this condition, an appropriate cell-marking process is required to discriminate each cell with a variation in light intensity. However, it is not possible to apply this procedure using a conventional microscopic technique. Therefore, a novel boundary construction operation was designed that can describe boundaries based not only on cell shapes but also on the light intensity of cells. The results of the operation enable us to quantify important characteristics, such as cell size distribution and perimeter size. The proposed method has advantages over the conventional segmentation (i.e., watershed) method in quantifying cell structures with larger cells; however, a longer computing time is required.