Using the histogram of oriented gradients for detecting cephalometric landmarks

The issue of locating cephalometric landmarks in skull radiography is necessary for orthodontic treatment, study of skeletal changes, staggered teeth diagnosis, etc. Currently, the steps to identify cephalometric landmarks are primarily based on experiences and the visual assessment of trained doctors. This paper proposes a technique which automatically identifies these landmarks on skull radiography using Histogram of Oriented Gradients (HOG) descriptor. At first, a set of photo is created to classify the regions containing desired benchmarks with other regions. These images will be performed by the HOG descriptor. Then the Support Vector Machine (SVM) algorithm will classify these images and create a reference model. Through this model, we have experimented on several skull radiographies of man at the age from 10 to 32 and women from 10 to 35. The result completely convinces as not only detects cephalometric landmarks on the sharpness of the images but also on the blurred images.