Gaetano Impoco 1, Laura Tuminello 1, and Giuseppe Licitra 1,2
1- CoRFiLaC – Consorzio Ricerca Filiera Lattiero-Casearia, Ragusa, Italy
2- D.A.C.P.A., University of Catania, Italy
Supervised learning approaches to image segmentation receive considerable interest due to their power and exibility. However, the training phase is not painless, often being long and tedious. Accurate image labelling can take several hours of expert operators’ valuable time. User interfaces are often speci cally designed to assist the user for the task at hand. This is clearly unfeasible for most application domains. We propose a simple segmentation framework based on classi cation and supervised incremental learning. A statistical model of pixel classes is learnt by incrementally adding new sample image patches to automatically-learned probability functions. Learning is iterated and re ned in a number of steps rather than executed in an off-line training phase. We show that off-line training and incremental labelling tend to produce similar statistical models, as the number of iterations grows. Comparable classi cation results are thus obtained with considerable less human effort.
Keywords: Segmentation, Incremental learning, Microscopy
Gaetano Impoco
caccamo@corfilac.it
