AI

Adaptive Fused Prior Transfer for Controllable Generative Image Compression

Read full article at source ↗

Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 06/10/2026, 18:18:25 UTC.

Briefings

A new controllable image compression codec named Adaptive Fused Prior Transfer for Controllable Generative Image Compression (AFP-GIC) transfers an adaptive fused prior from a frozen pretrained AdaCode model. This approach reduces decoder latency by 18.1% and uses 31.10 million fewer inference parameters compared to DC-VIC, while maintaining competitive PSNR and SSIM on benchmark datasets.

Key points

• AFP-GIC transfers an adaptive fused prior from a frozen pretrained AdaCode model to enable controllable reconstruction without transmitting the prior itself.
• The codec achieves 18.1% lower decoder latency and uses 31.10 million fewer inference parameters than DC-VIC under the unified benchmark.

Context

Existing controllable codecs generally rely on single-codebook reconstruction priors, whereas AFP-GIC transfers an adaptive fused prior from a frozen pretrained AdaCode model. This method guides latent formation via encoder-side features and predicts compatible priors from compressed representations and control variables.

Limitations

The study uses a unified benchmark for comparison; specific results for NIQE and visual comparisons are noted as having the clearest perceptual gains, though exact NIQE values are not provided in the evidence.

Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 23/10/2026, 03:03:10 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.

View archived briefing on IPFS
bafybeigfymaqa6v5hcjxw7j4w6yu7jkqlrfbfkexl5e2lzpfz6z7i3d4g4