Our main contribution is to use contrastive learning for creating image-like embeddings suitable for registration, and

We thank the reviewers for their thorough evaluation. Apart from comparing with [29], we also mention [22] and [25] Table 3] for the biomedical dataset) we believe this is feasible. The reviews contain many suggestions on how to clarify and improve the article. The main computational cost of the method is linear w.r.t. We thank the reviewer for advising us to explore gCCA, which seems highly relevant.