No, the paper means primarily the weight sharing in each kernel filter within each convolutional layer, and the stacking of these layers in deep networks.