On Perceptual Lossy Compression: The Cost of Perceptual Reconstruction and An Optimal Training Framework

Zeyu Yan, Fei Wen, Rendong Ying, Chao Ma, Peilin Liu
Proceedings of the 38th International Conference on Machine Learning, PMLR 139:11682-11692, 2021.

Abstract

Lossy compression algorithms are typically designed to achieve the lowest possible distortion at a given bit rate. However, recent studies show that pursuing high perceptual quality would lead to increase of the lowest achievable distortion (e.g., MSE). This paper provides nontrivial results theoretically revealing that, 1) the cost of achieving perfect perception quality is exactly a doubling of the lowest achievable MSE distortion, 2) an optimal encoder for the “classic” rate-distortion problem is also optimal for the perceptual compression problem, 3) distortion loss is unnecessary for training a perceptual decoder. Further, we propose a novel training framework to achieve the lowest MSE distortion under perfect perception constraint at a given bit rate. This framework uses a GAN with discriminator conditioned on an MSE-optimized encoder, which is superior over the traditional framework using distortion plus adversarial loss. Experiments are provided to verify the theoretical finding and demonstrate the superiority of the proposed training framework.

Cite this Paper

BibTeX
@InProceedings{pmlr-v139-yan21d, title = {On Perceptual Lossy Compression: The Cost of Perceptual Reconstruction and An Optimal Training Framework}, author = {Yan, Zeyu and Wen, Fei and Ying, Rendong and Ma, Chao and Liu, Peilin}, booktitle = {Proceedings of the 38th International Conference on Machine Learning}, pages = {11682--11692}, year = {2021}, editor = {Meila, Marina and Zhang, Tong}, volume = {139}, series = {Proceedings of Machine Learning Research}, month = {18--24 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v139/yan21d/yan21d.pdf}, url = {https://proceedings.mlr.press/v139/yan21d.html}, abstract = {Lossy compression algorithms are typically designed to achieve the lowest possible distortion at a given bit rate. However, recent studies show that pursuing high perceptual quality would lead to increase of the lowest achievable distortion (e.g., MSE). This paper provides nontrivial results theoretically revealing that, 1) the cost of achieving perfect perception quality is exactly a doubling of the lowest achievable MSE distortion, 2) an optimal encoder for the “classic” rate-distortion problem is also optimal for the perceptual compression problem, 3) distortion loss is unnecessary for training a perceptual decoder. Further, we propose a novel training framework to achieve the lowest MSE distortion under perfect perception constraint at a given bit rate. This framework uses a GAN with discriminator conditioned on an MSE-optimized encoder, which is superior over the traditional framework using distortion plus adversarial loss. Experiments are provided to verify the theoretical finding and demonstrate the superiority of the proposed training framework.} }
Endnote
%0 Conference Paper %T On Perceptual Lossy Compression: The Cost of Perceptual Reconstruction and An Optimal Training Framework %A Zeyu Yan %A Fei Wen %A Rendong Ying %A Chao Ma %A Peilin Liu %B Proceedings of the 38th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2021 %E Marina Meila %E Tong Zhang %F pmlr-v139-yan21d %I PMLR %P 11682--11692 %U https://proceedings.mlr.press/v139/yan21d.html %V 139 %X Lossy compression algorithms are typically designed to achieve the lowest possible distortion at a given bit rate. However, recent studies show that pursuing high perceptual quality would lead to increase of the lowest achievable distortion (e.g., MSE). This paper provides nontrivial results theoretically revealing that, 1) the cost of achieving perfect perception quality is exactly a doubling of the lowest achievable MSE distortion, 2) an optimal encoder for the “classic” rate-distortion problem is also optimal for the perceptual compression problem, 3) distortion loss is unnecessary for training a perceptual decoder. Further, we propose a novel training framework to achieve the lowest MSE distortion under perfect perception constraint at a given bit rate. This framework uses a GAN with discriminator conditioned on an MSE-optimized encoder, which is superior over the traditional framework using distortion plus adversarial loss. Experiments are provided to verify the theoretical finding and demonstrate the superiority of the proposed training framework.
APA
Yan, Z., Wen, F., Ying, R., Ma, C. & Liu, P.. (2021). On Perceptual Lossy Compression: The Cost of Perceptual Reconstruction and An Optimal Training Framework. Proceedings of the 38th International Conference on Machine Learning, in Proceedings of Machine Learning Research 139:11682-11692 Available from https://proceedings.mlr.press/v139/yan21d.html.

Related Material