A 2D wavelet-based multiscale approach with applications to the analysis of digital mammograms

A wavelet-based multifractal spectrum (MFS) for the analysis of images that possess an erratically changing oscillatory behavior at various scales is constructed and estimated. The methodology is applied to the analysis of mammograms. The key contribution is that the analysis is not focused on microcalcifications, but on the background of the image, thus presenting a new modality to be combined with other diagnostic tools. Differences in image backgrounds between malignant and normal cases are found, in terms of multifractal descriptors. The new tool is compared with another spectral method, based on monofractal descriptors.1