Open AccessAn Appearance-Based Method for Parametric Video Registration
Author(s) -
Xavier Orriols,
Lluis Barceló,
Xavier Binefa
Publication year - 2003
Publication title -
elcvia. electronic letters on computer vision and image analysis
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.15
H-Index - 11
ISSN - 1577-5097
DOI - 10.5565/rev/elcvia.62
Subject(s) - subspace topology , transformation (genetics) , parametric statistics , representation (politics) , frame (networking) , sequence (biology) , artificial intelligence , computer vision , computer science , parametric model , optical flow , mathematics , reference frame , algorithm , image (mathematics) , pattern recognition (psychology) , statistics , telecommunications , biochemistry , chemistry , genetics , politics , biology , political science , law , gene
In this paper we address the problem of multi frame video registration using the combination of an appearance-based technique and a parametric model of the transformations. This technique uses an image that is selected as reference frame, and therefore, estimates the transformation that occurred to each frame in the sequence respect to this absolute referenced one. Both global and local information are employed to the estimation of these registered images. Global information is applied in terms of linear appearance subspace constraints, under the subspace constancy assumption [4], where variabilities of each frame respect to the reference frame are encoded. Local information is used by means of a polynomial parametric model that estimates the velocities field evoluton in each frame. The objective function to be minimized considers both issues at the same time, i.e., the appearance representation and the time evolution across the sequence. This function is the connection between the global coordinates in the subspace representation and the time evolution and the parametric optical flow estimates. Thus, the appearance constraints result to take into account al the images in a sequence in order to estimate the transformation parameters