Research Seminar - June 10, 2002

Outlier Detection in Video Sequences under Affine Projection

Abstract:

A method for outlier detection in structure and motion recovery for the affine camera will be presented. It is an extension of the well-known Tomasi-Kanade factorization technique designed to handle outliers, and can also be seen as an importation of the LMedS technique or RANSAC into the factorization framework. Based on the computation of distances between subspaces, it relates closely with the subspace-based factorization methods for the perspective case presented by Sparr and others and the subspace-based factorization technique for the affine camera with missing data by Jacobs. Key features of the method presented here are its ability to compare different subspaces, the derivation of the relationship between subspace distances and reprojection errors, and the complete automation of the processes of detection and elimination of outliers. Its performance and effectiveness are demonstrated by experiments involving simulated and real video sequences.