English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Deep networks have achieved great success in image rescaling (IR) task thatseeks to learn the optimal downscaled representations, i.e., low-resolution(LR) images, to reconstruct the original high-resolution (HR) images. Comparedwith super-resolution methods that consider a fixed downscaling scheme, e.g.,bicubic, IR often achieves significantly better reconstruction performancethanks to the learned downscaled representations. This highlights theimportance of a good downscaled representation in image reconstruction tasks.Existing IR methods mainly learn the downscaled representation by jointlyoptimizing the downscaling and upscaling models. Unlike them, we seek toimprove the downscaled representation through a different and more direct way:optimizing the downscaled image itself instead of the down-/upscaling models.Specifically, we propose a collaborative downscaling scheme that directlygenerates the collaborative LR examples by descending the gradient w.r.t. thereconstruction loss on them to benefit the IR process. Furthermore, since LRimages are downscaled from the corresponding HR images, one can also improvethe downscaled representation if we have a better representation in the HRdomain. Inspired by this, we propose a Hierarchical Collaborative Downscaling(HCD) method that performs gradient descent in both HR and LR domains toimprove the downscaled representations. Extensive experiments show that our HCDsignificantly improves the reconstruction performance both quantitatively andqualitatively. Moreover, we also highlight the flexibility of our HCD since itcan generalize well across diverse IR models.

Downscaled Representation Matters: Improving Image Rescaling with Collaborative Downscaled Images