Indentifying parts of images that "stand out", and do so reliably, has been a long-standing goal of computer vision. Recent attempts to do this include the SIFT and MSER operators. At the UEA we have developed stable salient contours (SSCs) which use processors know as sieves. MSERs, which are one of the best known saliency operators, also are derived from sieves so we find that SScs have similar properties. SSCs can be used for image retrieval and machine learning.
Dr Richard Harvey, Dr. Yuxuan Lan, Mr. Nick Smith, Mr. Roberto Perez-Torres