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

The recent use of diffusion prior, enhanced by pre-trained text-image models,has markedly elevated the performance of image super-resolution (SR). Toalleviate the huge computational cost required by pixel-based diffusion SR,latent-based methods utilize a feature encoder to transform the image and thenimplement the SR image generation in a compact latent space. Nevertheless,there are two major issues that limit the performance of latent-baseddiffusion. First, the compression of latent space usually causes reconstructiondistortion. Second, huge computational cost constrains the parameter scale ofthe diffusion model. To counteract these issues, we first propose a frequencycompensation module that enhances the frequency components from latent space topixel space. The reconstruction distortion (especially for high-frequencyinformation) can be significantly decreased. Then, we propose to useSample-Space Mixture of Experts (SS-MoE) to achieve more powerful latent-basedSR, which steadily improves the capacity of the model without a significantincrease in inference costs. These carefully crafted designs contribute toperformance improvements in largely explored 4x blind super-resolutionbenchmarks and extend to large magnification factors, i.e., 8x image SRbenchmarks. The code is available at https://github.com/amandaluof/moe_sr.

Image Super-resolution Via Latent Diffusion: A Sampling-space Mixture Of Experts And Frequency-augmented Decoder Approach