CONTENT-ADAPTIVE IMAGE SUPER-RESOLUTION USING TWO-STAGE CLASSIFICATION

초록

This paper proposes an image super-resolution algorithm that improves the high frequency synthesis by using accurate classification for low-resolution patches, and also provides smaller memory size as well as less computational cost than the exiting methods. Experimental results show that the proposed algorithm provides better subjective visual quality as well as objective quality. Thus, the proposed algorithm is appropriate for real-time image scaling for digital consumer applications.