On the evolution of subcritical regions for the Perona-Malik equation