Tomakeuseofthepriorknowledgeoftheimagemoreeffectivelyandrestoremoredetailsoftheedgesandstructures,anovelsparsecodingobjectivefunctionisproposedbyapplyingtheprincipleofthenon-localsimilarityandmanifoldlearningonthebasisofsuper-resolutionalgorithmviasparserepresentation.Firstly,thenon-localsimilarityregularizationtermisconstructedbyusingthesimilarimagepatchestopreservetheedgeinformation.Then,themanifoldlearningregularizationtermisconstructedbyutilizingthelocallylinearembeddingapproachtoenhancethestructuralinformation.Theexperimentalresultsvalidatethattheproposedalgorithmhasasignificantimprovementcomparedwithseveralsuper-resolutionalgorithmsintermsofthesubjectivevisualeffectandobjectiveevaluationindices.
