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

Learned image compression methods have shown superior rate-distortionperformance and remarkable potential compared to traditional compressionmethods. Most existing learned approaches use stacked convolution orwindow-based self-attention for transform coding, which aggregate spatialinformation in a fixed range. In this paper, we focus on extending spatialaggregation capability and propose a dynamic kernel-based transform coding. Theproposed adaptive aggregation generates kernel offsets to capture validinformation in the content-conditioned range to help transform. With theadaptive aggregation strategy and the sharing weights mechanism, our method canachieve promising transform capability with acceptable model complexity.Besides, according to the recent progress of entropy model, we define ageneralized coarse-to-fine entropy model, considering the coarse globalcontext, the channel-wise, and the spatial context. Based on it, we introducedynamic kernel in hyper-prior to generate more expressive global context.Furthermore, we propose an asymmetric spatial-channel entropy model accordingto the investigation of the spatial characteristics of the grouped latents. Theasymmetric entropy model aims to reduce statistical redundancy whilemaintaining coding efficiency. Experimental results demonstrate that our methodachieves superior rate-distortion performance on three benchmarks compared tothe state-of-the-art learning-based methods.

Dynamic Kernel-Based Adaptive Spatial Aggregation for Learned Image Compression