Technical paper on measuring local change for image segmentation. Number 1.
Williams, P.S., Mike Alder · UWA Profiles and Research Repository (UWA) · 1997
This technical paper examines the possible measures for identifying regions of an image which are highly homogeneous. In previous work the Euclidean distance between local feature vectors with scaled components is used to obtain this measurement and in turn to identify the initial segments for a subsequent segmentation algorithm. Although effective, the ability to reliably locate less homogeneous regions, particularly textured ones, requires improvement.