In this paper, we present an algorithm to estimate the cell counts from cultured rat B104 neuroblastoma cell images. Assuming that cells are alike, the algorithm identifies the representative cell based on processing the sorted size sequence of the threshold-segmented regions. The size of the representative cell is used to estimate the number of cells in each cluster where cells are attached and inseparable. Preprocessing procedures include the homomorphic filtering for improving the evenness of the image background, and gray-level morphologic dilation and erosion operations for filling the hollow cells. Results on the B104 neuroblastoma cell images are provided and compared with the numbers derived from manual cell count by pathologists.