Robust Factorization Methods Using a Gaussian/Uniform Mixture Model
International Journal of Computer Vision
An Iterative Multiresolution Scheme for SFM with Missing Data
Journal of Mathematical Imaging and Vision
3-D retinal curvature estimation
IEEE Transactions on Information Technology in Biomedicine - Special section on body sensor networks
| Hi-index | 0.00 |
It is widely known that, for the affine camera model, both shapeand motion can be factorized directly from the so-called imagemeasurement matrix constructed from image point coordinates. Theability to extract both shape and motion from this matrix by asingle SVD operation makes this shape-from-motion approachattractive; however, it cannot deal with missing feature pointsand, in the presence of outliers, a direct SVD to the matrix wouldyield highly unreliable shape and motion components. In this paper,we present an outlier correction scheme that iteratively updatesthe elements of the image measurement matrix. The magnitude andsign of the update to each element is dependent upon the residualrobustly estimated in each iteration. The result is that outliersare corrected and retained, giving improved reconstruction andsmaller reprojection errors. Our iterative outlier correctionscheme has been applied to both synthesized and real videosequences. The results obtained are remarkably good.