Fractal based criteria to evaluate the performance of digital image magnification techniques

Abstract

The ability of a magnification technique to preserve the fractal dimension of an image after magnification is proposed as a measure to compare the performances of the different digital image magnification techniques. Fractal dimensions are attractive measures for such comparisons since they are not only data structure dependent but also invariant to scale changes in the original data. Tests with image data sets indicate that bilinear interpolation preserves the fractal dimension up to 8 times magnification while bicubic interpolation preserves it up to 4 times.