Passive method for 3D reconstruction of human jaw: theory and application.

Oral dental applications based on visual data pose various challenges. There are problems with lighting (effect of saliva, tooth dis-colorization, gum texture, and other sources of specularity) and motion (even inevitable slight motions of the upper/ lower jaw may lead to errors far beyond the desired tolerance of sub-millimeter accuracy). Nowadays, the dental CAM systems have become more compromised and accurate to obtain the geometric data of the jaw from the active sensor (laser scanner). However, they have not met the expectations and the needs of dental professionals in many ways. The probes in these systems are bulky { even their newer versions - and are hard to maneuver. It requires multiple scans to get full coverage of the oral cavity. In addition, the dominant drawback of these systems is the cost. Stereo-based 3D reconstruction provides the highest accuracy among vision systems of this type. However, the evaluation of it\u27s performance for both accuracy results and the number of 3D points that are reconstructed would be affected by the type of the application and the quality of the data that is been acquired from the object of interest. Therefore, in this study, the stereo-based 3D reconstruction will vi be evaluated for the dental application. The handpiece of sensors holder would reach to areas inside the oral cavity, the gap between the tooth in the upper jaw and the tooth in the lower jaw in these areas would be very small, in such the stereo algorithms would not be able to reconstruct the tooth in that areas because of the distance between the optical sensors and the object of interest \tooth as well as the configuration of optical sensors are contradicted the geometric constraint roles of the stereo-based 3D reconstruction. Therefore, the configuration of the optical sensors as well as the number of sensors in the hand piece of sensors holder will be determined based on the morphological of the teeth surfaces. In addition to the 3D reconstruction, the panoramic view of a complete arch of human teeth will be accomplished as an application of dental imaging. Due to the low rate of features on teeth surfaces, the normal tooth surface is extracted using shape from shading. The extracted surface normals impact many imprecise values because of the oral environment; hence an algorithm is being formulated to rectify these values and generate normal maps. The normal maps reveal the impacted geometric properties of the images inside an area, boundary, and shape. Furthermore, the unrestricted camera movement problem is investigated. The camera may be moved along the jaw curve with different angles and distances due to handshaking. To overcome this problem, each frame is tested after warping it, and only correct frames are used to generate the panoramic view. The proposed approach outperforms comparing to the state-of-art auto stitching method