Multi-image gradient-based algorithms for motion estimation

Image registration algorithms based on gradient methods provide quantitativ e motion measurements from sequences of video images. Although such measure ments can be degraded by image noise, larger degradations typically result from systematic bias in the algorithms that is present even if the images a re noise-free. To improve the accuracy of motion measurements, we develop a new class of multi-image algorithms based on multidimensional digital filt ers. The new algorithms provide better estimates of spatial and temporal gr adients and also compensate for motion blur caused by the nonzero acquisiti on time of the imager. We optimize filters to measure arbitrary motions, an d we illustrate the results when those filters are used to estimate constan t velocity movements. We also show results for filters that are optimized f or harmonic analysis of periodic motions. Using these algorithms, systemati c bias in the amplitude of sinusoidal motion is less than 0.001 pixels for motions smaller than 1 pixel in amplitude. This represents a hundredfold de crease in bias compared to existing methods. (C) 2001 society of Photo-Opti cal Instrumentation Engineers.