New Techniques for Tracking Sequences of Digitized Images.

摘要

In Chapter 1, the research objectives are outlined, the image tracking task is defined, and the component parts of a generalized image-tracking system are discussed. Chapter 2 presents as background material a summary of applications for image tracking systems, and some classical similarity detection techniques. In Chapter 3, a two-dimensional signal-to-noise ratio is developed for minimum norm similarity detectors and an upper bound is developed and experimentally verified for the probability of error associated with a particular type of misregistration. Chapter 4 presents three techniques for enhancing the signal-to-noise ratio, Chapter 5 presents an adaptive Kalman filter which is shown to be very helpful in reducing misregistration errors, and Chapter 6 develops and analyzes the performance of an integrated tracking algorithm which incorporates a complementary set of component algorithms. Chapter 7 provides a summary of this research and some suggestions for future work.