One-to-Many Network for Visually Pleasing Compression Artifacts Reduction
This work addresses the issue of overly smooth results in image compression artifact reduction, which is important for applications requiring high visual fidelity, though it appears incremental as it builds on existing deep learning approaches.
The paper tackled the problem of compression artifacts reduction in images by proposing a one-to-many network that uses perceptual, naturalness, and JPEG losses to improve visual quality, resulting in dramatic visual improvements over state-of-the-art methods.
We consider the compression artifacts reduction problem, where a compressed image is transformed into an artifact-free image. Recent approaches for this problem typically train a one-to-one mapping using a per-pixel $L_2$ loss between the outputs and the ground-truths. We point out that these approaches used to produce overly smooth results, and PSNR doesn't reflect their real performance. In this paper, we propose a one-to-many network, which measures output quality using a perceptual loss, a naturalness loss, and a JPEG loss. We also avoid grid-like artifacts during deconvolution using a "shift-and-average" strategy. Extensive experimental results demonstrate the dramatic visual improvement of our approach over the state of the arts.