3D-model-based tracking offers one possibility to explicate the manner in which spatial coherence can be exploited for the analysis of image sequences. Two seemingly different ap...
Hendrik Dahlkamp, Artur Ottlik, Hans-Hellmut Nagel
This paper proposes an appearance generative mixture model based on key frames for meanshift tracking. Meanshift tracking algorithm tracks object by maximizing the similarity betwe...
In this paper we propose a new probabilistic relaxation framework to perform robust multiple motion estimation and segmentation from a sequence of images. Our approach uses displa...
Bayesian methods for visual tracking model the likelihood of image measurements conditioned on a tracking hypothesis. Image measurements may, for example, correspond to various fi...
* We present a novel approach for grouping from motion, based on a 4-D Tensor Voting computational framework. From sparse point tokens in two frames we recover the dense velocity f...