The motion estimation methods are used for modeling of various physical processes, the behavior of objects, and prediction of events. In this chapter the moving objects in videos are generally considered. Such motion estimation methods are classified as comparative methods and gradient methods. The comparative motion estimation methods are usually used in real-time applications. Many aspects of block-matching modifications are discussed including Gaussian mixture model, Lie operators, bilinear deformations, multi-level motion model, etc. The gradient motion estimation methods assist to realize the motion segmentation in complex dynamic scenes because only they provide a required accuracy. Application of the 2D tensors (in spatial domain) or the 3D tensors (in spatio-temporal domain) depends from the solved problem. Development of the gradient motion estimation methods is necessary for intelligent recognition of objects and events in complex scenes, video indexing in multimedia databases. © Springer-Verlag Berlin Heidelberg 2012.
CITATION STYLE
Favorskaya, M. (2012). Motion estimation for objects analysis and detection in videos. Intelligent Systems Reference Library, 29, 211–253. https://doi.org/10.1007/978-3-642-24693-7_8
Mendeley helps you to discover research relevant for your work.