Statistical reconstruction and feature tracking of temporally irregular data sequences

For irregular observation periods of the sea surface, the multi-filter system has been developed to reconstruct the series of observed images for regular time intervals by recovering missing measurements and dynamically interpolating sequential observations over time. A general spatial structure of the image is represented by an autoregressive response model and a polynomial model is employed for dynamic interpolation to track the underlying variation over time. This approach allows successive refinement of the structure of objects that are barely detectable in the observed series, using an expectation maximum likelihood algorithm. In this study, missing data are recovered by using an interpolation function and a quad-tree pyramid structure, and an alternative approach for feature tracking is proposed on the basis of local temporal trend