How to convolution and deconvolution share weight

How to convolution and deconvolution share weights?

Does this occur in a particular paper?