Summary
We present an approach for probabilistic contour prediction within the framework of an object tracking system. We combine level-set methods for image segmentation with optical flow estimations based on probability distribution functions (pdfs) calculated at each image position. Unlike most recent level-set methods that consider exclusively the sign of the level-set function to determine an object and its background, we introduce a novel interpretation of the value of the level-set function that reflects the confidence in the contour. To this end, in a sequence of consecutive images, the contour of an object is transformed according to the optical flow estimation and used as the initial object hypothesis in the following image. The values of the initial level-set function are set according to the optical flow pdfs and thus provide an opportunity to incorporate the uncertainties of the optical flow estimation in the object contour prediction.
See the full content of this document
Extract
A Probabilistic Prediction Method for Object Contour Tracking
(ProQuest: ... denotes formulae omitted.)
1. IntroductionIn this paper we propose an object contour tracking approach based on level-set methods for image segmentation and correlation-based patch-matching methods for optical flow estimation. Using level-set methods for object detection enables us to overcome the problems imposed by non-rigid object deformations and object appearance changes. In tracking applications with dynamic template adaption these changes lead to template drift, and in applications without template adaption to a decreased robustness. Utilizing probabilistic optical flow for the prediction of the object contour constitutes a non-parametric prediction model that is capable of representing non-rigid object deformation as well as complex and rapid object movements, thus providing a segmentation method with a reliable initial contour that leads to a robust and quick convergence of the level-set method even in the presence of a comparably low camera frame rate. Furthermore, we introduce a novel interpretation of the value of the level-set function. Unlike most recent levelset methods that consider exclusively the sign of the level-set function to determine an object and its surroundings, we use the value of the level-set function to reflect the confidence in the predicted initial contour. This yields a robust and quick convergence of the level-set method in those sections of the contour with a high initial confidence and a flexible and mostly unconstrained (and thus also quick) convergence in those sections with a low initial confidence.In the field of image segmen...See the full content of this document
Sponsored links
