I have to match my noisy image with the closest matching clean image (say template). To find the closest match between my image and match image which of the meta-learning approach should I select. I am thinking about the Siamese or matching network/prototypical network. I think in Siamese I have to match my noisy image with all templates to find the correct match. So do you think if I can use a matching network and produce a better result?

Note- During training, I will train my data with some template and their corresponding noisy image but at the time of the test new templates can be added on which model is not trained. Do you think a matching network will still work for this case?