Segmentation of motion in an image sequence is one of the most challenging problems in image processing, while at the same time one that finds numerous applications. To date, a wealth of approaches to motion segmentation have been proposed. Many of them suffer from the local nature of models used. Global models, such as those based on Markov random fields, perform, in general, better. In this paper, we propose a new approach to motion segmentation that is based on a global model. The novelty of the approach is twofold. First, inspired by recent work of other researchers we formulate the problem as that of region competition, but we solve it using the level set methodology. The key features of a level set representation, as compared to active contours, often used in this context, are its ability to handle variations in the topology of the segmentation and its numerical stability. The second novelty of the paper is the formulation in which, unlike in many other motion segmentation algori...

We propose a pattern classification based approach for simultaneous 3-D object modeling and segmentation in image volumes. The 3-D objects are described as a set of overlapping ellipsoids.

We employ a prediction model for moving object velocity and location estimation derived from Bayesian theory. The optical flow of a certain moving object depends on the history of its previous values. A joint optical flow estimation and moving object segmentation algorithm is used for the initialization of the tracking algorithm. The segmentation of the moving objects is determined by appropriately classifying the unlabeled and the occluding regions. Segmentation and optical flow tracking is used for predicting future frames.

This paper presents an iterative maximum likelihood framework for motion segmentation via the pairwise checking of pixel blocks. We commence from a characterisation of the motion blocks in terms of a matrix of pairwise similarity weghts for their motion vectors. The eigenvectors of this similarity weight matrix represent the initial pairwise clusters, i.e the independant motions present in the scene. We develop a maximum likelihood framework which allows to update both the link weight matrix and the associated set of pairwise clusters. We experiment with the resulting clustering method on a number of real world motion sequences. Here ground truth data indicates that the method can result in motion classification errors as low as 3%. 1

this paper, we shall describe robust and adaptive training algorithms that have been developed the past three years and aim at enhancing the capabilities of the self-organizing and the RBF neural networks [3]-[12]

This paper describes a new method to detect moving objects in a dynamic scene based on background subtraction. The main goal of the method is to obtain and keep a stable background image to cope with variations on environmental changing conditions. In this way, we use a double background (long-term background and short-term background) to deal with temporal stability and fast changes. In addition, this method computes the temporal changes in the video sequence by a local convolution mask taking into account the information of the pixel neighborhood, being less sensitive to noise. Besides, the method classifies the regions of change in moving and static blobs. The first ones represent real moving objects, and the second are related to illumination changes and noise. Finally, experimental results and a performance measure establishing the confidence of the method are presented.

Emerging applications like asynchronous distant learning and collaborative engineering require organization of media streams as multimedia presentations. The browsing of presentations enables interactive surfing of the multimedia documents. We propose spatiotemporal browsing of multimedia presentations in the sense that browsing can be performed both in the spatial and temporal domain.

Abstract. This paper provides a comparison study among a set of robust diffusion algorithms for processing optical flows. The proposed algorithms combine the smoothing ability of the heat kernel, modelled by the local Hessian, and the outlier rejection mechanisms of robust statistics algorithms. Smooth optical flow variation can be modelled very well using heat kernels. The diffusion kernel is considered Gaussian, where the covariance matrix implements the inverse of the local Hessian. Robust statistics operators improve the results provided by the heat kernel based diffusion, by rejecting outliers and by avoiding optical flow oversmoothing. Alpha-trimmed mean and median statistics are considered for robustifying diffusion kernels. The robust diffusion smoothing is applied onto multiple frames and is extended to 3D lattices. 1

The image sequence is represented as a set of moving regions which make up moving objects. Motion, position and graylevel (or color) information is used for segmenting the moving objects. A criterion is proposed for modeling the 3-D motion and segmentation. After identifying the occluding regions, the moving objects are tracked over the next frames. Prediction is employed for estimating the future moving object position and its optical flow. 1 Introduction Various approaches have been proposed for optical flow estimation and motion segmentation [1]. The maximization of the a posteriori probability has been considered in [2]. A classification approach was proposed for jointly segmenting the moving objects and their corresponding optical flow in [3]. Median Radial Basis Function (MRBF) algorithm which relies on robust statistics was employed for estimating the moving object characteristic vectors [3]. We provide a classification based criterion for the 3-D segmentation of the image sequ...

Previous work has shown that information from different artificial vision approaches to the same problem can be combined to produce more robust results. Often, information from a technique looking at a completely different aspect of an image can also be of use. This paper reports the use of fuzzy set theory to combine the results of image processing techniques for different problems. Classical decision theory is used to guide the choice of fuzzy methods to take account of observation inter-dependencies and risk in a meaningful way. The approach is illustrated by the use of texture information to improve the results of motion estimation methods. Introduction For many image processing problems, there exist a number of different techniques whose relative performance is dependent on the particular image or image sequence used. Furthermore, the solution to one image processing problem may contain information which is of use in a different problem. It makes sense to combine the individual ...