Crafting a Toolchain for Image Restoration by Deep Reinforcement Learning
This addresses image restoration for applications like photography or medical imaging, but is incremental as it builds on existing reinforcement learning and network toolbox concepts.
The paper tackles image restoration by developing RL-Restore, a method that uses reinforcement learning to dynamically select small specialized networks from a toolbox, resulting in more parameter-efficient restoration of images with complex distortions compared to conventional networks.
We investigate a novel approach for image restoration by reinforcement learning. Unlike existing studies that mostly train a single large network for a specialized task, we prepare a toolbox consisting of small-scale convolutional networks of different complexities and specialized in different tasks. Our method, RL-Restore, then learns a policy to select appropriate tools from the toolbox to progressively restore the quality of a corrupted image. We formulate a step-wise reward function proportional to how well the image is restored at each step to learn the action policy. We also devise a joint learning scheme to train the agent and tools for better performance in handling uncertainty. In comparison to conventional human-designed networks, RL-Restore is capable of restoring images corrupted with complex and unknown distortions in a more parameter-efficient manner using the dynamically formed toolchain.