Authors:
Mingxing Hu
1
;
David J. Hawkes
1
;
Graeme P. Penney
2
;
Daniel Rueckert
3
;
Philip J. Edwards
4
;
Fernado Bello
4
;
Michael Figl
3
and
Roberto Casula
5
Affiliations:
1
Centre for Medical Image Computing, University College London, United Kingdom
;
2
Department of Imaging Sciences, King’s College London, United Kingdom
;
3
Department of Computing, Imperial College, United Kingdom
;
4
Department of Surgical Oncology and Technology, Imperial College, United Kingdom
;
5
Cardiothoracic Surgery, St. Mary’s Hospital, United Kingdom
Keyword(s):
Video Mosaicing, Robotic Assisted Minimally Invasive Surgery, Homography, Trifocal Tensor, Bundle Adjustment.
Related
Ontology
Subjects/Areas/Topics:
Image Processing
;
Informatics in Control, Automation and Robotics
;
Robotics and Automation
;
Vision, Recognition and Reconstruction
Abstract:
Constructing a mosaicing image with broader field-of-view has become an interesting topic in image guided diagnosis and treatment. In this paper, we present a robust method for video mosaicing in order to provide more guiding information for robotic assisted minimally invasive surgery. Outliers involved in the feature dataset are removed using trifocal constraints, homographies between images are estimated with -norm optimization and chained together in a practical way. Finally refinement based on bundle adjustment is applied to minimize the error between reprojection and feature measurement. The proposed method has been tested with endoscopic images from Totally Endoscopic Coronary Artery Bypass (TECAB) surgery. The results showed our method performs better than other typical methods in terms of accuracy and robustness to deformation.