One long argument: Azriel Rosenfeld and the genesis of modern image systems Virginio Cantoni a , Stefano Levialdi b, * a Dipartimento di Informatica e Sistemistica, Universita `degli Studi, Pavia, Italy b Dipartimento di Informatica, Universita `di Roma, La Sapienza, Roma, Italy Received 9 July 2004; received in revised form 6 October 2004 Abstract We present our vision on the history of multiprocessor architectures for image processing seen under the light of DarwinÕs Theory of Evolution, remembering our fruitful and entertaining scientific meetings with Professor Azriel Rosenfeld along the nearly 30years of our acquaintance. We experienced the blooming of novel systems (70s and 80s), the harvesting of such systems (90s) and later, the extinction of the massively parallel systems in favor of the adop- tion of large clusters of fast commercial processors to accomplish the same image processing tasks. Ó 2004 Published by Elsevier B.V. Keywords: Parallel image processing; Computer vision architectures; Multiprocessor systems evolution; Natural selection In memoriam One of us has known Professor Azriel Rosen- feld for many years since 1968 at the Tirrenia School in Italy of Pattern and Character Recogni- tion, where he lectured on basic algorithms for im- age processing. Since then, we have met at the different International Conferences and Work- shops that were held, most of them under his influ- ence, in different parts of the world. He has been, for the whole Italian community, a scientific guide not only to the literature but also, and most of all, to the exciting subjects which were at the cutting edge of image processing research. His critical evaluation of our research lines together with a definite sense of humor, have helped us to design new architectures and algorithms, along the many years of our acquaintance. Among the many sys- tem projects that originated at Maryland Univer- sity under his inspiration, Azriel also conceived and published a novel massive architecture (PRISM) alas, so massive, that remained a paper machine. At the Rome International Conference on Pattern Recognition, ICPR held on 1988, we 0167-8655/$ - see front matter Ó 2004 Published by Elsevier B.V. doi:10.1016/j.patrec.2004.10.015 * Corresponding author. E-mail addresses: virginio.cantoni@unipv.it (V. Cantoni), levialdi@di.uniroma1.it (S. Levialdi). Pattern Recognition Letters 26 (2005) 277–285 www.elsevier.com/locate/patrec