Moving object detection and trajectory estimation in the transform/spatiotemporal mixed domain

A novel discrete transform/spatiotemporal mixed domain, or MixeD, moving object detection and trajectory estimation algorithm is introduced. It is known that the energy of a linear trajectory moving object is confined to a plane in the 3-D frequency domain. The 2-D discrete Fourier transform (DFT) is combined with a 1-D high-resolution spectral estimation method to identify and locate the plane, and thereby estimate the object´s trajectory. An example is provided to demonstrate that the MixeD moving object detection and trajectory estimation algorithm is able to discriminate several moving objects in an image sequence, and to provide accurate estimates of their trajectories