Despite the high performance achieved in goniometers-control for tomographic tilt series recording, image alignment is mandatory to obtain accurate reconstructions. In addition, due to the high resolution currently expected from tomograms, the existing alignment methods (based on cross-correlation, fiducial-markers and landmarks) needs to be improved. Among the alternative methods presently proposed for general purpose image registration, Scale Invariant Feature Transform (SIFT) [1] has demonstrate its performance for the detection of common features occurring in several images. SIFT have been successfully used for panoramic stitching [2], object detection [1] and proposed as a method for image alignment in electron tomography [3].
We have implemented SIFT and two other feature-based registration algorithms porposed since the publication of SIFT (ORB [4], BRISK [5]) in TomoJ [6] and evaluated their respective performances on tilt series data from different transmission electron microscopy (TEM) acquisition methods (TEM, EFTEM, STEM, and Cryo-TEM). We will present here the results obtained validating the use of feature-based registration algorithms on tilt series from the different acquisition modes.
To cite this abstract:
Amandine Verguet, Sylvain Trepout, Sergio Marco, Cédric Messaoudi; Evaluation of feature-based registration algorithms for the improvement of tilt-series alignment in electron tomography. The 16th European Microscopy Congress, Lyon, France. https://emc-proceedings.com/abstract/evaluation-of-feature-based-registration-algorithms-for-the-improvement-of-tilt-series-alignment-in-electron-tomography/. Accessed: December 2, 2023« Back to The 16th European Microscopy Congress 2016
EMC Abstracts - https://emc-proceedings.com/abstract/evaluation-of-feature-based-registration-algorithms-for-the-improvement-of-tilt-series-alignment-in-electron-tomography/