Shape and motion from image streams under orthography: a factorization method
International Journal of Computer Vision
Performance of optical flow techniques
International Journal of Computer Vision
Hierarchical mixtures of experts and the EM algorithm
Neural Computation
Statistical physics, mixtures of distributions, and the EM algorithm
Neural Computation
Recursive 3-D Visual Motion Estimation Using Subspace Constraints
International Journal of Computer Vision
ICCV '95 Proceedings of the Fifth International Conference on Computer Vision
Object-oriented coding using successive motion field segmentation and estimation
ICIP '95 Proceedings of the 1995 International Conference on Image Processing (Vol. 1)-Volume 1 - Volume 1
A Paraperspective Factorization Method for Shape and Motion Recovery
A Paraperspective Factorization Method for Shape and Motion Recovery
On Convergence Properties of the EM Algorithm for Gaussian Mixtures
On Convergence Properties of the EM Algorithm for Gaussian Mixtures
Hi-index | 0.00 |
This paper presents a framework for object-oriented scenesegmentation in video, which uses motion as the major characteristic todistinguish different moving objects and then to segment the scene intoobject regions. From the feature block (FB) correspondences through at leasttwo frames obtained via a tracking algorithm, the reference featuremeasurement matrix and feature displacement matrix are formed. We propose atechnique for initial motion clustering of the FBs, in which the principalcomponents (PC) of the two matrices are adopted as the motion features. Themotion features have several advantages: (1) They are low-dimensional(2-dim). (2) They preserve well both the spatial closeness and the motionsimilarity of their corresponding FBs. (3) They tend to form distinctiveclusters in the feature space, thus allowing simple clustering schemes to beapplied. The Expectation-Maximization (EM) algorithm is applied forclustering the motion features. For those scenes involving mainly the cameramotion, the PC-based motion features will exhibit nearly parallel lines inthe feature space. This facilitates a simple and yet effective layerextraction scheme. The final motion-based segmentation involves labeling ofall the blocks in the frame. The EM algorithm is again applied to minimizean energy function which takes motion consistency andneighborhood-sensitivity into account. The proposed algorithm has beenapplied to several test sequences and the simulation results suggest apromising potential for video applications.