This paper presents scale invariant feature transform (SIFT) -based camera motion estimation for bronchoscope tracking. We show a method for predicting bronchoscope motion that uses SIFT features to obtain inter-frame position and orientation displacements. We improve the performance of bronchoscopic tracking by employing image registration initialized by the output of feature-based camera motion prediction. Furthermore, the proposed method is evaluated on real bronchoscopic video data and phantom data. Experimental results from both datasets demonstrate a significant performance boost of tracking without an additional position sensor.     «

This paper presents scale invariant feature transform (SIFT) -based camera motion estimation for bronchoscope tracking. We show a method for predicting bronchoscope motion that uses SIFT features to obtain inter-frame position and orientation displacements. We improve the performance of bronchoscopic tracking by employing image registration initialized by the output of feature-based camera motion prediction. Furthermore, the proposed method is evaluated on real bronchoscopic video data and phant...     »