Image restoration and enhancement is a classical research area in image processing. Previously, adaptive local and nonlocal approximations have been popular. Local approximations attempt to estimate the image content in a locally adaptive neighbourhood. Nonlocal methods exploit the self-similarity within the whole image without the constraint of locality. The former tends to be more efficient and the latter would produce better results. Recently, learning-based techniques adopting advances in machine learning and computer vision, such as sparse coding and dictionary learning, have attracted much more attention and been applied to image/video restoration and enhancement. These techniques can represent image contents better using learned dictionaries. In addition, some novel application areas, e.g., legacy photos and paintings, HD/3D displays, mobile and portable devices, and web-scale data, have prompted new research interests in image/video restoration and enhancement.