N. Grammalidis, G. Nikolakis, M. G. Strintzis. Use of classification for fast fractal-based coding in the wavelet domain. In Proc. of the Conference Fractals in Engineering, Arcachon, June 1997.
Based on the standard fractal transformation in the spatial domain, simple relations may be found relating coefficients in detail (high-pass) subbands in the wavelet domain. In this paper three classification approaches are examined so as to reduce the computationsl complexity of the search procedure used for selecting the "optimal" domain block corresponding to each range block. The first is based on scale-invariant features of the range and domain blocks, specifically scale-invariant moments. the second is an adaptation of a classification method originally proposed for classifying blocks in the spatial domain by ordering the mean average luminances of the four block quadrants. The third method stems from the minimization of the mean square error corresponding to the optimal scale factor. While all three techniques vield better results than a simple random search for the optimal domain block, the results of the third approach are clearly superior.
@InProceedings{GrNiSt97,
Author = {Grammalidis, N. and Nikolakis, G. and Strintzis, M. G.},
Title = {Use of classification for fast fractal-based coding in the wavelet domain},
BookTitle = {Proc. of the Conference Fractals in Engineering},
Address = {Arcachon},
Month = {June},
Year = {1997}
}