Active Control of Resolution for Stable Visual Tracking

Success of visual tracking typically relies on the ability to process visual information su ciently fast. Often a dynamic systemmodel of target motion is used to estimate the target location within the image and a region of interest (ROI) is used to reduce the amount of image data processing. This has proven e ective, provided the ROI is su ciently large to detect the target and su ciently small to be processed quick enough. Here we formally consider the size of the ROI and the resolution of the ROI to ensure that tracking is stable. Dynamic system formulation of visual tracking usually requires speci cation of the dynamics of the target. We analyze motions which can be described by linear time-invariant dynamical systems (although the image motion may be highly non-linear). One can successfully analyze the required ROI size and resolution to ensure stable tracking.