[Spm] Measuring fractal dimension for small image/profile size

[Sergei V. Kalinin]

Hello everybody
    I am looking for advice regarding estimates of fractal 
dimensionality and multifractal f(alpha) spectra from SPM images and 
profiles [e.g. perimeters for thresholded images] with small number of 
pixels. In many cases, direct use of box-counting method (even if the 
partitions are shifted to improve statistics) yields systematic errors 
due to small range of accessible bin sizes [e.g., even checkerboard 
becomes fracta!]. It looks that for some cases the analysis can be 
improved if the morphology of the features is known (e.g. G. Ouillon et 
al, J. Geophys. Res. 5477 (1996)). I am wondering if there are 
references addressing this issue for arbitrary feature morphology.
    Thank you in advance
    Sergei