TY - GEN
T1 - ON DISENTANGLED TRAINING FOR NONLINEAR TRANSFORM IN LEARNED IMAGE COMPRESSION
AU - Li, Han
AU - Li, Shaohui
AU - Dai, Wenrui
AU - Cao, Maida
AU - Kan, Nuowen
AU - Li, Chenglin
AU - Zou, Junni
AU - Xiong, Hongkai
N1 - Publisher Copyright:
© 2025 13th International Conference on Learning Representations, ICLR 2025. All rights reserved.
PY - 2025
Y1 - 2025
N2 - Learned image compression (LIC) has demonstrated superior rate-distortion (RD) performance compared to traditional codecs, but is challenged by training inefficiency that could incur more than two weeks to train a state-of-the-art model from scratch. Existing LIC methods overlook the slow convergence caused by compacting energy in learning nonlinear transforms. In this paper, we first reveal that such energy compaction consists of two components, i.e., feature decorrelation and uneven energy modulation. On such basis, we propose a linear auxiliary transform (AuxT) to disentangle energy compaction in training nonlinear transforms. The proposed AuxT obtains coarse approximation to achieve efficient energy compaction such that distribution fitting with the nonlinear transforms can be simplified to fine details. We then develop wavelet-based linear shortcuts (WLSs) for AuxT that leverages wavelet-based downsampling and orthogonal linear projection for feature decorrelation and subband-aware scaling for uneven energy modulation. AuxT is lightweight and plug-and-play to be integrated into diverse LIC models to address the slow convergence issue. Experimental results demonstrate that the proposed approach can accelerate training of LIC models by 2 times and simultaneously achieves an average 1% BD-rate reduction. To our best knowledge, this is one of the first successful attempt that can significantly improve the convergence of LIC with comparable or superior rate-distortion performance. Code will be released at https://github.com/qingshi9974/AuxT.
AB - Learned image compression (LIC) has demonstrated superior rate-distortion (RD) performance compared to traditional codecs, but is challenged by training inefficiency that could incur more than two weeks to train a state-of-the-art model from scratch. Existing LIC methods overlook the slow convergence caused by compacting energy in learning nonlinear transforms. In this paper, we first reveal that such energy compaction consists of two components, i.e., feature decorrelation and uneven energy modulation. On such basis, we propose a linear auxiliary transform (AuxT) to disentangle energy compaction in training nonlinear transforms. The proposed AuxT obtains coarse approximation to achieve efficient energy compaction such that distribution fitting with the nonlinear transforms can be simplified to fine details. We then develop wavelet-based linear shortcuts (WLSs) for AuxT that leverages wavelet-based downsampling and orthogonal linear projection for feature decorrelation and subband-aware scaling for uneven energy modulation. AuxT is lightweight and plug-and-play to be integrated into diverse LIC models to address the slow convergence issue. Experimental results demonstrate that the proposed approach can accelerate training of LIC models by 2 times and simultaneously achieves an average 1% BD-rate reduction. To our best knowledge, this is one of the first successful attempt that can significantly improve the convergence of LIC with comparable or superior rate-distortion performance. Code will be released at https://github.com/qingshi9974/AuxT.
UR - https://www.scopus.com/pages/publications/105010264646
M3 - 会议稿件
AN - SCOPUS:105010264646
T3 - 13th International Conference on Learning Representations, ICLR 2025
SP - 84718
EP - 84738
BT - 13th International Conference on Learning Representations, ICLR 2025
PB - International Conference on Learning Representations, ICLR
T2 - 13th International Conference on Learning Representations, ICLR 2025
Y2 - 24 April 2025 through 28 April 2025
ER -