Efficient image compression method improves speed and quality at low bitrates
Efficient Generative Image Compression with Quantized Latents Refinement
Computer Vision and Pattern Recognition
Summary
Compressing images efficiently while keeping them looking good is tricky, especially when trying to make the process fast. The authors noticed that simple compression can squash small details together, losing fine features in images. They created a new approach, called GR-GIC, that improves image quality by refining the compressed information and speeds up the process using a better way to handle data. This method works well on standard tests and runs quickly even on less powerful devices.
What this means in practice
- •For mobile app developers: Integrate this fast and high-quality image compression method to reduce data usage and improve user experience when sharing images at low bitrates.
- •For cloud storage providers: Use the method to store more images in less space while maintaining perceptual quality, saving storage costs and bandwidth.
Authors
Hao Cao, Yinhuan Huang, Wenqi Guo, Zhijin Qin, Jungong Han
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.