Estimation of Deformation Degree with Uncertainty and Missing Data

In this paper we propose an approach to estimating the deformation degree of a non-rigid shape from monocular image sequence with uncertainty and missing data. The non-rigid shape can be represented as a weighted combination of K rigid basis shapes. Since K denotes the number of basis shapes that can represent the shape sequence, it provides a measure of deformation degree of the shape sequence. K is computed from the tracked positions of a sequence of non-rigid shapes by using rank analysis method. It is well known that the uncertainty of tracked position depends on the underlying spatial intensity structure in the image, which has strong directionality. Hence, the factor of directional uncertainty is taken into account in our deformation degree estimation method. In order to cope with the missing data problem, ¿-ß-¿ filter-based visual tracking is proposed to track the motion of feature points between the consecutive image frames and recover the missing points. Experimental results from real image sequences demonstrate the feasibility of the proposed approach.