An accurate, automatic method for markerless alignment of electron tomographic images

Qi Chu,Fa Zhang,Kai Zhang,Xiaohua Wan,Mingwei Chen,Zhiyong Liu
DOI: https://doi.org/10.1109/BIBM.2010.5706597
2010-01-01
Abstract:Accurate alignment of electron tomographic images without using embedded gold particles as fiducial markers is still a challenge. Here we propose a new markerless alignment method that employs Scale Invariant Feature Transform features (SIFT) as virtual markers. It differs from other types of feature in a way the sufficient and distinctive information it represents. This characteristic makes the following feature matching and tracking steps automatic and more reliable, which allows for estimating alignment parameters accurately. Furthermore, we use Sparse Bundle Adjustment (SPA) with M-estimation to estimate alignment parameters for each image. Experiments show that our method can achieve a reprojection residual less than 0.4 pixel and can approach the same accuracy of marker alignment. Besides, our method can apply to adjusting typical misalignments such as magnitude divergences or in-plane rotation and can detect bad images.
What problem does this paper attempt to address?