New Algorithms for Variable Time Delay and Nonuniform Image Motion Estimation

This text introduces and investigates estimation algorithms for signal registration. Signal registration can be defined as the problem of estimating a varying displacement between two random processes. This problem has great importance in many applications, perhaps the most familiar of which lies in the field of passive sonar. The bearing of the signal source is related to the time delay (displacement) between two random waveforms. In the general signal registration problem, the time delay is itself a random function of time. Part one of this book is devoted to the problem of variable time delay estimation. Another application of the signal registration lies in image processing where registration is called motion estimation. By determining the relative displacement (motion) between the intensities of consecutive image frames, one can encode an image more efficiently. In this application, the displacement is a position-dependent vector process. Part two discusses the problem of nonuniform image motion estimation.