Chen Ning and Clifford A. Reiter, Best analogs for replacing missing image data, Computers & Graphics 31 (2007) 617-624.

Identifying the historical data that is the best analog with a pattern from which a forecast is sought allows time series data to be extrapolated. That technique of best analogs is most effective when the data contains underlying deterministic chaos. Here we apply similar techniques, modified to use two space dimensions instead of one time dimension, to fill-in and extrapolate missing image data. The technique is successful at replacing significant amounts of missing data with reasonable data derived from the image itself.

See also: · Preprint, and some images, animations related to the paper [Best Analogs Materials] · Time Series Fractal Forecasting [Abstract]