A sparse-modeled ROI for GLAM construction in image classification problems—A case study of breast cancer

Abstract

Image segmentation is a process to determine regions of interest (ROI) in mammograms. Mammograms can be classified by extracting textural features of ROI using Gray Level Aura Matrices (GLAM). Scientists are selecting a fixed window size for all ROIs to find respective GLAM, though the masses will not occur in regular two-dimensional geometries. This paper makes an attempt to replicate the problem but by choosing arbitrary shape of masses as they occur. It is found that this kind of natural selection of the arbitrary shape yielded drastic reduction in time complexity by adopting the method suggested by us.