Improving The Reconstruction Quality by Overfitted Decoder Bias in Neural Image Compression

Jourairi, Oussama; Balcilar, Muhammet; Lambert, Anne; Schnitzler, François

doi:10.1109/pcs56426.2022.10018052

Cited by 3 publications

(1 citation statement)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In literature, the proposed methods to reduce this gap are three folds. The first ones involved latent finetuning for the given image [15][16][17][18][19], the second ones model parameters finetuning [20][21][22][23][24] for better performance on a given single image or the re-parameterization of the entropy model in order to better fit the latents without fine-tuning the model [25]. In practice, only the first class of method allows to keep the same decoder and does not increase the decoder complexity.…”

Section: Introductionmentioning

confidence: 99%

Entropy Coding Improvement for Low-complexity Compressive Auto-encoders

Galpin¹,

Balcilar²,

Lefèbvre³

et al. 2023

Preprint

View full text Add to dashboard Cite

End-to-end image and video compression using auto-encoders (AE) offers new appealing perspectives in terms of rate-distortion gains and applications. While most complex models are on par with the latest compression standard like VVC/H.266 on objective metrics, practical implementation and complexity remain strong issues for real-world applications. In this paper, we propose a practical implementation suitable for realistic applications, leading to a low-complexity model. We demonstrate that some gains can be achieved on top of a state-of-the-art low-complexity AE, even when using simpler implementation. Improvements include off-training entropy coding improvement and encoder side Rate Distortion Optimized Quantization. Results show a 19% improvement in BDrate on basic implementation of fully-factorized model, and 15.3% improvement compared to the original implementation. The proposed implementation also allows a direct integration of such approaches on a variety of platforms.

show abstract