Con preghiera di diffusione tra tutti i possibili interessati, scusandomi per invii multipli. Cordialmente, Giacomo Aletti
===============
Nell'ambito del Seminario di Matematica Applicata, in collaborazione col centro ADAMSS e il CIMAB, il giorno giovedì 12 Maggio 2016, alle ore 14.30, nella Sala di Rappresentanza (piano terra) del Dipartimento di Matematica dell'Universita' degli Studi di Milano, Via C. Saldini, 50, Milano,
"Image segmentation, graph clustering and Hough transform methods for object detection and measurement in images" Luca CALATRONI Dipartimento di Matematica Universita' di Genova
Abstract: In this talk we consider the problem of image segmentation through the minimisation of a Ginzburg-Landau-type functional defined on graphs (i.e. the pixel images). Such approach has first been considered by A. Bertozzi and A. Flenner and provides a binary segmentation of the image via the extraction and the comparison of features (RGB, texture,...) with a dictionary in terms of an appropriate similarity measure. From a mathematical point of view, the segmentation is obtained by exploiting the spectral properties of the differential operators appearing when taking the $\ell^2$ gradient flow of the functional. In order to overcome the numerical difficulties due to the large size of the images considered, Nyström matrix completion techniques and convex splitting methods are employed. We apply such method to the problem of scale detection in images where a fuzzy region of interest is present together with a measurement tool (e.g. a ruler). In particular, by means of a Hough transform based algorithm, we apply our combined method to several measurement tasks arising in real-world applications such as zoology, medicine and archaeology.