Dictionary learning based multitask image restoration

In recent years, there have been increased interests in the study of dictionary learning (DL). Learned dictionaries lead to state-of-the-art results in image processing. Hence, many new DL methods were presented. This paper proposes a novel DL model and an algorithm to solve this model. We call the proposed algorithm problem-guided dictionary learning (PG-DL). PG-DL can deal with many problems in image processing. Taking noised image inpainting and removing mixed noise as examples, the experiments show that the PG-DL can describe the image content effectively and leads to valid performance.