Is there such a thing as a convolution with non-uniform dilation?

Deep Learning
1

For example, a kernel that spreads out over an image first by one unit, next by 2 units, next 4 units?

Thanks for any references or leads.