A new approach for characterizing man-made and natural objects from two-dimensional data has been developed. The approach is based on the use of a fractal signature, which is a measure of local fractal dimension as a function of scale and space. Estimates of the fractal signature are obtained by using computationally efficient morphological filters. Consistency of the fractal signature and the effects of edges and anomalies are examined by using real and synthetic data. Preliminary results using optical and infrared data demonstrate the potential of using fractal signatures for object characterization and discrimination.
© 1990 Optical Society of America
T. Peli, "Multiscale fractal theory and object characterization," J. Opt. Soc. Am. A 7, 1101-1112 (1990)