We suggest a computational approach to the orbit segmentation for computer tomography (CT) images. The first step was to perform Hounsfield unit thresholding to segment the bony structure. Then a three dimensional (3D) mesh model was generated. Poisson surface reconstruction was applied to the screened vertices that lay on the inner orbital walls. These procedures effectively interpolate the broken surfaces due to orbital fissures; various nerves foramina; and thin bone structure around the orbit. A validation was performed on the CT images of a dried skull with dental impression material filled orbits. The volume differences are 2.2 and 1.5% respectively for the two orbits. Surface differences are within ±0.46mm RMS. The differences are not clinically significant. The main novelty of proposed method is the ability in selecting vertices according to the desired orientation and therefore it is robust against broken structure in the segmentation.