Abstarct: The domain of medical analysis has improved rapidly in recent years due to the availability and affordability of digital radiography imaging equipment and associated algorithms. Furthermore, there has been significant activity in the automation of the medical diagnostic process. One of such processes is cephalometric analysis that includes cephalometric landmark detection and analysis of linear and angular relations between these landmarks. This process is manually intensive. Although, there are some computer software programs that can perform various analysis after determining landmark positions on the input image, these programs could not locate landmarks on the input image automatically and leave landmark detection step to users to do it manually by some tools these programs provide for them. This thesis describes an approach, baxsed on image processing techniques and machine learning, to automate the landmark detection process. A cephalometric analysis involves locating a number of points in an X-ray image and determining the linear and angular relationships between them. If the points can be located accurately enough, the rest of the analysis is straightforward. The investigated steps were as follows: Firstly, an object detection algorithm was employed as a domain independent approach to find a first estimation of the landmark’s position. The approach was tested on a selection of landmarks, ranging from easy to very difficult. This method used histograms of oriented gradient besides support vector machines. Although, we were able to estimate the location of whole 16 landmarks in more than 84 percent of cases, there was still room for improvement. Next, in the second part, some improvements over parameters of selected features and classifier were done. Also, some methods for choosing the best position for the first estimation of the landmark has been investigated and a new feature introduced to improve this estimation results. The results were acceptable for first estimation of landmark’s position. In the third part, some attempts have been done to pinpoint the exact location of landmarks around the region estimated in the previous attempts. The accuracy of landmark detection was promising for some landmarks but for others in cluttered background seemed to need some improvements. Therefore, we tried to improve our results in the next part. In this part, the overall process has been reviewed and changes made to improve the overall performance of the system. These improvements include changing SVM kernel type and applying image enhancement algorithms before computing feature vectors from the input image. It can be concluded from the results that the proposed system could find location of landmarks such as Gnathion، Menton، Sella، Porion، Pogonion and Upper Incisor Tip in a distance of 2mm from their real location with a precision more than 65 percent. It should be noticed that the accuracy of the system for locating the other landmarks is comparable with other works in this filed but it should be improved for clinical use in cephalometry. The major outcome of this work is that the method described in this thesis could be used as the basis of an automated system. The orthodontists would be required to manually correct a few errors before completing the analysis.