Abstract: An automated cephalogram image analysis method is disclosed. In this method, a step is first performed for building a reference database in which a set of tracing feature curve models respectively representing a set of reference cephalometric patterns are established based on a set of tracing records. Then, a step of pattern comparison and analysis is performed. In this step, at least one cephalogram is first input, and then a step is performed for comparing the image of the cephalogram with the tracing feature curve models, thereby selecting at least one cephalometric pattern and at least one feature curve model belonging to the cephalometric pattern from the reference cephalometric patterns and the tracing feature curve models. Thereafter, the feature curve model is fitted to the image of the cephalogram so as to obtain at least one tracing of the at least one cephalogram.