Segmentación automática de billetes mexicanos basada en un modelo de color y referencias geométricas

Recently, the development of aiding systems for visually impaired people, based on image processing and object recognition, has attracted interest. In this work, a method for automatic segmentation of Mexican banknotes based on color and geometric references (points of interest obtained automatically) is presented. Our objective is to precisely identify the banknote region in order to extract color information useful for the later classification of the banknote. The method consists of: 1) automatically draw a region of reference; 2) generate a color model and a similarity image affected by a geometric weight function; 3) thresholding of the similarity image to produce the final segmentation. The method was evaluated on a set of 926 images of banknotes obtained under 97 Research in Computing Science 114 (2016) pp. 97–110; rec. 2016-03-17; acc. 2016-05-14 different illumination; the method showed to be fast and achieve high performance.