Structure-Preserving Image Super-resolution via Contextualized Multi-task Learning | Read Paper on Bytez