Automatic registration of printed analog screen-film mammograms
Mario Muštra, Danijel Mileta, Mislav Grgić · International Conference on Systems, Signals and Image Processing · 2011
Computer aided detection (CAD) in mammography relies on accurate image segmentation. If we consider using CAD on non-digital mammograms, one of the steps which need to be performed is scanning. After scanning, images have to be segmented and this is more complicated than segmentation of digital mammograms. In this paper we propose a method for automatic registration of printed and scanned mammograms. This process allows efficient manual segmentation which gives as accurate as possible results. To translate manual segmentation results performed by radiologists from paper to digital format, image needs to be registered. Manual image registration generally takes too much time and therefore is not as efficient as automatic. This paper considers automatic registration of scanned images using Hough transform. Accurate detection and alignment of marker lines on scanned images gives the correct angle of rotation, as well as translation and resizing coefficient.