Architecture for fractal dimension estimation based on Minkowski-Bouligand method using integer distances
Isabela Rossales, Maximiliam Luppe · 2016
Fractal dimension is an extremely important tool in shape analysis and characterization, including tasks from signal processing to image processing. One of the reasons for such great interest is the power of the fractal dimension to properly express the intricacy and self-similarity of signals. One of the best methods to obtain the fractal dimension is the Minkowski-Bouligand. However, this method is computationally exhaustive due to the use of the Euclidean distance transform, which involves floating point number calculation. In this work, we present a dedicated hardware implementation proposal, based on the Minkowski-Bouligand method with integer distances and suitable for reconfigurable devices. The implementation is tested on an image database of plant leaves, yielding satisfactory results.