Fuzzy segmentation of the breast region in mammograms
This thesis is an investigation of segmenting the breast region in mammogram by using fuzzy image processing. Mammography is currently accepted as the most effective imaging for detecting breast cancer. This thesis addresses the fundamental question of segmenting breast region from background. Four main theoretical approaches, used to separate the breast area from mammograms, are: histogram based thresholding, gradient edge-based methods, region-based techniques, and connectivity-based segmentations. A fuzzy clustering algorithm is theoretically considered as one of most suitable for delineating the skin-air interface in mammogram, based on the difficulty of breast segmentation. The combination of the edge direction of breast contour with the standard FCM improves the segmentation result in breast segmentation. A robust segmentation module, which attempts to blend currently available image processing techniques into a single segmentation task, is used with FCM plus edge direction, in order to obtain the better segmentation result.