Edge Detection in Dynamic Vision

Using a model of an edge's motion through a sequence of images, the problem of its recognition can be formulated as a stochastic filtering problem. The Extended Kalman Filter for such a system is considered in detail and is shown to be interpretable as a sequence of oriented spatial convolutions. Preliminary results show that the edge localization obtained using this filter is substatinally better than that obtained using the Sobel operator on each image individually. time varying imagery, such as obtained from AGVs or motion'''', there is a temporal coherency in the sequence of images that can be exploited in the design of filters with a temporal as well as spatial basis. By capturing this temporal coherency with a dynamical model, it is possible to obtain an Extended Kalman Filter which localizes edges. This is the purpose of the work reported in this paper. Related work is reported in >>) where Extended Kalman Filters are used to combine multiple pieces of data about higher order entities such as three dimensional points, lines and planes.