Identification of masses in mammograms by image sub-segmentation

Abstract

Mass detection in mammography is a complex and challenge problem for digital image processing. Partitional clustering algorithms are a good alternative for automatic detection of such elements, but have the disadvantage of having to segment an image into a number of regions, the number of which is unknown in advance, in addition to discrete approximations of the regions of interest. In this work we use a method of image sub-segmentation to identify possible masses in mammography. The advantage of this method is that the number of regions to segment the image is a known value so the algorithm is applied only once. Additionally, there is a parameter ? that can change between 1 and 0 in a continuous way, offering the possibility of a continuous and more accurate approximation of the region of interest. Finally, since the identification of masses is based on the internal similarity of a group data, this method offers the possibility to identify such objects even from a small number of pixels in digital images. This paper presents an illustrative example using the traditional segmentation of images and the sub-segmentation method, which highlights the potential of the alternative we propose for such problems. � 2011 Springer-Verlag Berlin Heidelberg.