Matrix factorization techniques for recommender systems given test set contains items not in train set

Given a movie-user rating dataset and a task to predict the user rating for a movie. I would like to know how matrix factorization techniques (collaborative filtering) would perform for when the movies present in the test set is not present in training set. I presume the features uncovered by matrix factorization in these scenarios would be bad because the movie was unseen during the training phase. Are there any other set of algorithms that deal with these kinds of scenarios?