Newest papers

Efficient Generative Image Compression with Quantized Latents Refinement

Hao Cao, Yinhuan Huang, Wenqi Guo, Zhijin Qin, Jungong Han · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.29691 · cited by 0 works (OpenAlex) · open-access PDF

Indonesian

GIC memberikan kualitas persepsi yang kuat pada tingkat bit rendah. Namun, metode GIC dengan kualitas persepsi yang kuat sering bergantung pada model difusi pretrained dengan latency tinggi, sementara metode GIC ringan baru-baru ini mengurangi latency tetapi memberikan kualitas persepsi yang lebih rendah. Kami memperhatikan bahwa, untuk metode GIC ringan dengan tingkat bit rendah, kuantizasi deterministik kasar dapat menghancurkan latent encoder dekat yang sesuai dengan rincian halus ke titik yang sama, membuat rincian ini sulit untuk dipulihkan. Bermotivasi oleh ini, kami mengusulkan GR-GIC, GIC yang efisien yang memberikan kinerja rekonstruksi kompetitif dengan inferensi cepat. Untuk kinerja yang lebih baik, kami memperkenalkan Guided Refinement (GR), yang berhasil menggabungkan latent refinement yang sesuai dengan konteks entropy panduan dengan gangguan adaptif untuk regulasi pel

Machine translation (Meta M2M100); the English original follows.

Abstract

Generative image compression (GIC) delivers strong perceptual quality at low bitrates. However, GIC methods with strong perceptual quality often rely on pretrained diffusion models with high latency, while recent lightweight GIC methods reduce latency but deliver lower perceptual quality. We observe that, for lightweight GIC methods at low bitrates, coarse deterministic quantization can collapse nearby encoder latents corresponding to different fine details to the same point, making these details difficult to recover. Motivated by this, we propose GR-GIC, an efficient GIC that delivers competitive reconstruction performance with fast inference. For better performance, we introduce Guided Refinement (GR), which combines latent refinement guided by entropy context with adaptive perturbations for training regularization, jointly improving reconstruction at low bitrates. For fast inference, we introduce reorganized projected residual vector quantization (P-RVQ), which equivalently reorganizes residual updates into low-dimensional lookups to expose greater parallelism and reduce sequential computation. Experiments on Kodak show that for performance, compared with OneDC, GR-GIC achieves BD-rate savings of 23.19%, 37.71%, 22.30%, and 22.14% for LPIPS, DISTS, PSNR, and FID, respectively. For fast inference, the highly optimized implementation of GR-GIC encodes and decodes in 4.60 and 5.98 ms on an NVIDIA A100. Specifically, the reorganized P-RVQ implementation achieves a $3.14\times$ average speedup over the original implementation under the same FP32 setting. On an NVIDIA Jetson Orin NX, GR-GIC achieves latency comparable to lightweight GIC methods.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Computer science, Algorithm, Residual, Speedup, Encoder, Data compression

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange