In dynamic and opportunistic access for cognitive radios, the leading trend is to split the available spectrum into separated channels and make the scavenging users access the ones considered free. However, it is likely that even the "occupied" ones do not use up all the resource but leave enough room for other connections. In this paper we present a method that handles the spectrum sensing with the tools of image processing. In detail, by means of the Wigner-Ville transform we are able to analyze the time and frequency features of the received interference, and to represent them as a greyscale image in which brightness reveals the spectrum occupancy. Then, through a proper mask definition and matching algorithm we shape a signal that fills up the "spectrum holes" and exploits the spatial dimension offered by the multiple antennas in order to set up a connection satisfying a specified quality requirement (i.e., the transmission rate). © 2010 IEEE.