Failure of L2 closure of the global BV domain
ID: failure-of-l2-closure-of-the-global-bv-domain
For , the radial function belongs to and has finite distributional total variation seminorm on a domain, but is not in . Cutting it off outside radius gives bounded-variation space functions tending to in , with variation tending to that of : the added jump costs . Thus the penalty equal to TV on and infinity elsewhere is not sequentially lower semicontinuous in . The closed homogeneous bounded-variation space extension avoids this domain issue.
New to topics? Read the docs here!