A pattern-less medical augmented reality system for craniofacial applications

An augmented reality (AR) system usually needs placing a known pattern in the field of view of the imaging camera. According to the appearance of the pattern observed, the posture of the camera can be estimated. Then visual objects can be overlaid at the right place on the image. In this study, a medical AR system that uses a patternless approach is presented and the system is suitable in craniofacial applications such as facial plastic surgery. First, the proposed approach detects Harris corners within the facial area. A RANSAC-based framework integrating with KLT tracking, camera posture estimation and outlier correction are then performed. The framework shows the ability on dealing with the outliers caused by incorrect tracking. The proposed system was tested with a plastic phantom and the experimental results show that the AR display result is more stable with the RANSAC-based framework than using the KLT tracking alone.