The ability to detect cephalometric landmarks in x-ray images plays a crucial role in several medical applications, including orthodontics and maxillofacial surgery. The manual annotation of landmarks is time-consuming and prone to inaccuracy. These limitations can be overcome by building a system for automated landmark detection. Towards this goal, we propose a fully automated system for the task of soft-tissue landmark detection. The proposed method combines two different types of descriptors, Haar-like features and spatial features, to build a strong classifier using Adaboost technique. In order to evaluate the proposed method, we introduce a new dataset for soft-tissue landmark detection task along with two evaluation protocols to define detection rate. The first protocol represents the detection rate within mean radial error (MRE) while the second protocol represents the detection rate within a predefined confidence region R. Reported experiments demonstrate the superior performance of the proposed method compared to state-of-the-art.