Low-Rank CP Tensor Decomposition for Lightweight Image Super-Resolution
- 1College of Mathematics and Statistics, Kashi University, Kashi 844000, China
Abstract
Deep convolutional neural networks have substantially advanced single-image super-resolution, or SR, but their high parameter counts and computational redundancy hinder deployment on resource-constrained edge devices. Most existing compression methods apply pruning or low-rank truncation as a post hoc step after training. This “train-then-compress” pipeline is reported in the compression literature to disrupt the spectral coherence of the learned feature manifold and to require costly fine-tuning to restore fidelity. We study an alternative, training-time approach that re-parameterizes the convolutional kernel with CANDECOMP/PARAFAC (CP) decomposition into three low-rank sublayers, optimizing directly on a rank-constrained manifold rather than compressing after convergence. We interpret the forward pass as a heuristic realization, rather than a formal equivalence, of a proximal gradient-descent–ADMM update for a tensor nuclear-norm-regularized problem, which motivates the design and its hyperparameters without invoking expensive singular value decompositions. An analysis of parameter complexity and gradient flow justifies the choice of CP over Tucker and tensor-train decompositions. Experiments on DIV2K, Kodak24, BSD100, and Set12 show that the CP module yields a 0.11 M-parameter lightweight model at parity with a same-budget reduced residual baseline (at most 0.15 dB difference on every test set), sacrificing roughly half a decibel relative to a 5.2× larger residual baseline. Applied to a large-capacity SR backbone, the same module removes 76.7% of the parameters, from 13.28 M to 3.09 M, at a PSNR cost of only 0.11–0.20 dB. These results indicate that training-time CP decomposition is a simple and parameter-efficient route to compact SR models in the extreme-compression regime.
Related
- TokenComSR: Task-Sensitivity-Guided Token Communication for Wireless Image Super-Resolution
- Testing Gaussian process with applications to super-resolution
- Eigensolvers for polynomial roots and tensor decomposition
- Heteroskedastic Canonical Polyadic Tensor Decomposition
- TCA-Oriented Attractor Editing in a Discrete Memristive Rulkov Neuron with Application to Lightweight IoV Image Encryption
BibTeX
@article{guo2026lowrank,
title = {{Low-Rank CP Tensor Decomposition for Lightweight Image Super-Resolution}},
author = {Huan Guo and Yuekai Liu and Shirong Li},
journal = {Mathematics},
year = {2026},
month = nov,
doi = {10.3390/math14214049},
url = {https://www.mdpi.com/2227-7390/14/21/4049},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Low-Rank CP Tensor Decomposition for Lightweight Image Super-Resolution
AU - Huan Guo
AU - Yuekai Liu
AU - Shirong Li
PY - 2026
DA - 2026/11/05
JO - Mathematics
DO - 10.3390/math14214049
UR - https://www.mdpi.com/2227-7390/14/21/4049
AB - Deep convolutional neural networks have substantially advanced single-image super-resolution, or SR, but their high parameter counts and computational redundancy hinder deployment on resource-constrained edge devices. Most existing compression methods apply pruning or low-rank truncation as a post hoc step after training. This “train-then-compress” pipeline is reported in the compression literature to disrupt the spectral coherence of the learned feature manifold and to require costly fine-tuning to restore fidelity. We study an alternative, training-time approach that re-parameterizes the convolutional kernel with CANDECOMP/PARAFAC (CP) decomposition into three low-rank sublayers, optimizing directly on a rank-constrained manifold rather than compressing after convergence. We interpret the forward pass as a heuristic realization, rather than a formal equivalence, of a proximal gradient-descent–ADMM update for a tensor nuclear-norm-regularized problem, which motivates the design and its hyperparameters without invoking expensive singular value decompositions. An analysis of parameter complexity and gradient flow justifies the choice of CP over Tucker and tensor-train decompositions. Experiments on DIV2K, Kodak24, BSD100, and Set12 show that the CP module yields a 0.11 M-parameter lightweight model at parity with a same-budget reduced residual baseline (at most 0.15 dB difference on every test set), sacrificing roughly half a decibel relative to a 5.2× larger residual baseline. Applied to a large-capacity SR backbone, the same module removes 76.7% of the parameters, from 13.28 M to 3.09 M, at a PSNR cost of only 0.11–0.20 dB. These results indicate that training-time CP decomposition is a simple and parameter-efficient route to compact SR models in the extreme-compression regime.
SN - 2227-7390
ER -