| As approximately one in twelve women in the western world suffer from breast cancer at some time in their lives, screening programs based on mammography have been instituted in most of western countries. The processing of X-ray mammographic images is technically one of the most challenging applications of digital image processing, largely because of the high degree of variability associated with both normal and abnormal diagnostic patterns.; The advent of digitally-acquired mammograms offers the possibility of improvements in early breast cancer detection as well as readiness for digital image processing and teleradiology. Although digital mammography is still in the trial stage, computer-aided diagnosis with digitized mammographic images has shown its potential and advantages in breast cancer detection and management in terms of mammographic feature enhancement, computer-aided interpretation, and image archival.; The thesis presented herein addresses all of the above issues with region-based adaptive image processing techniques. First, a region-based adaptive contrast enhancement method is introduced for improving the visibility of mammographic features, and its potential of diagnostic performance improvement in early detection of breast cancer is demonstrated based on the two Receiver Operating Characteristics (ROC) analyses. Then, shape-based analysis of microcalcifications and asymmetric density patterns is presented and their effectiveness is demonstrated. Finally, lossless image data compression with a segmentation-based technique and an additional transformation-based technique to improve the Joint Bi-level Image experts Group (JBIG) coding technique are presented, and shown to outperform other advanced, popular lossless compression methods. |
