ADAPTIVE WAVELET TECHNIQUES FOR COMPRESSING DIGITAL MAMMOGRAMS

Abstract

In this work we investigate the possible benefit of employing adaptive wavelet algorithms instead of the classical fixed pyramidal wavelet decomposition for the compression of digital mammograms. In particular, we target on adaptive wavelet packet and NSMRA decompositions. We observe that information cost function optimized wavelet packet subband structures do not offer compression performance gain in this case whereas NSMRA decompositions moderately improve the results of classical wavelet decompositions. Due to the lack of fast and reliable search algorithms fixed NSMRA decompositions need to be generated and employed for classes of similar images.