Compression and enhancement of digital mammograms with wavelet footprints

We propose a near lossless representation of digital mammograms using the wavelet footprint basis, that can simultaneously enhance the diagnostic features of the mammogram and yield some compression. Our simulations on sample mammograms show that this representation can achieve upto a 4X reduction in the number of samples required. The mammograms reconstructed from this representation have Peak Signal to Noise Ratios (PSNR) of around 39 to 40dB on average. More significantly, important diagnostic features in the mammogram can be visually enhanced by a simple scaling of the footprint components of the mammogram.