Learning Single-Image Super-Resolution in the JPEG Compressed Domain

Fuente: arXiv
Enregistré dans:
Détails bibliographiques
Auteurs principaux: Srinivasan, Sruthi, Shakibapour, Elham, Rawther, Rajy, Saeedi, Mehdi
Format: Preprint
Publié: 2025
Sujets:
Accès en ligne:
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866915653203525632
author Srinivasan, Sruthi
Shakibapour, Elham
Rawther, Rajy
Saeedi, Mehdi
author_facet Srinivasan, Sruthi
Shakibapour, Elham
Rawther, Rajy
Saeedi, Mehdi
contents Deep learning models have grown increasingly complex, with input data sizes scaling accordingly. Despite substantial advances in specialized deep learning hardware, data loading continues to be a major bottleneck that limits training and inference speed. To address this challenge, we propose training models directly on encoded JPEG features, reducing the computational overhead associated with full JPEG decoding and significantly improving data loading efficiency. While prior works have focused on recognition tasks, we investigate the effectiveness of this approach for the restoration task of single-image super-resolution (SISR). We present a lightweight super-resolution pipeline that operates on JPEG discrete cosine transform (DCT) coefficients in the frequency domain. Our pipeline achieves a 2.6x speedup in data loading and a 2.5x speedup in training, while preserving visual quality comparable to standard SISR approaches.
format Preprint
id arxiv_https___arxiv_org_abs_2512_04284
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Learning Single-Image Super-Resolution in the JPEG Compressed Domain
Srinivasan, Sruthi
Shakibapour, Elham
Rawther, Rajy
Saeedi, Mehdi
Computer Vision and Pattern Recognition
Artificial Intelligence
Deep learning models have grown increasingly complex, with input data sizes scaling accordingly. Despite substantial advances in specialized deep learning hardware, data loading continues to be a major bottleneck that limits training and inference speed. To address this challenge, we propose training models directly on encoded JPEG features, reducing the computational overhead associated with full JPEG decoding and significantly improving data loading efficiency. While prior works have focused on recognition tasks, we investigate the effectiveness of this approach for the restoration task of single-image super-resolution (SISR). We present a lightweight super-resolution pipeline that operates on JPEG discrete cosine transform (DCT) coefficients in the frequency domain. Our pipeline achieves a 2.6x speedup in data loading and a 2.5x speedup in training, while preserving visual quality comparable to standard SISR approaches.
title Learning Single-Image Super-Resolution in the JPEG Compressed Domain
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2512.04284