Purpose: To evaluate the accuracy of detecting cephalometric landmarks automatically by computer. Materials and Methods: Digital image processing algorithms (edge-based and morphological) in addition to mathematical algorithms (knowledge-based) were used to develop a program for the identification of 15 lateral cephalometric landmarks on digital cephalograms. Results: The automatic landmark identification system was clinically acceptable (p > 0.05 comparing with 2-mm error range) for five landmarks. Conclusion: The automatic landmark identification is accurate for only some landmarks. Additional studies may be necessary to improve this method. Palabras clave: lateral cephalometry, software, automatic landmark identification