CCB_Meetings_ShantanuJoshi2007 < CCB

CCB Shantanu Joshi's Visit November 01, 2007.

Abstract:
- We present a framework for incorporating prior information about high-probability shapes in the process of contour extraction and object recognition in images. Here one studies shapes as elements of an infinite-dimensional, non-linear, quotient space, and statistics of shapes are defined and computed intrinsically using differential geometry of this shape space. Prior probability models are constructed on the tangent bundle of shape space. Past work on boundary extraction has used active curves driven by vector fields that were based on image gradients and roughness penalties. The proposed method incorporates a prior knowledge of shapes in the form of gradient fields in addition to the previously used image vector fields. Through experimental results, we demonstrate the use of prior shape models in estimation of object boundaries, and their success in handling partial obscuration and missing data. Furthermore, we will also describe the use of this framework in shape-based object recognition or classification. This Bayesian shape extraction approach is supposed to yield a significant improvement in detection of objects in presence of occlusions or obscurations.