View in EDS

Cascaded Sliding-Window-Based Relativistic GAN Fusion for Perceptual and Consistent Video Super-Resolution

Saved in:
Bibliographic Details
Title: Cascaded Sliding-Window-Based Relativistic GAN Fusion for Perceptual and Consistent Video Super-Resolution
Authors: Li, Dingyi
Contributors: School of Computer Science and Engineering Nanjing, Nanjing University of Science and Technology (NJUST), Zhongzhi Shi, Michael Witbrock, Qi Tian, TC 12
Source: IFIP Advances in Information and Communication Technology ; 6th International Conference on Intelligence Science (ICIS) ; https://inria.hal.science/hal-05142880 ; 6th International Conference on Intelligence Science (ICIS), Oct 2024, Nanjing, China. pp.232-247, ⟨10.1007/978-3-031-71253-1_17⟩
Publisher Information: CCSD
Springer Nature Switzerland
Publication Year: 2024
Subject Terms: Video Super-Resolution, Perceptual Quality, Temporal Consistency, Information Science, Intelligent Information Processing, [INFO]Computer Science [cs]
Subject Geographic: Nanjing, China
Description: Part 5: Perceptual Intelligence ; International audience ; Perceptual video super-resolution aims at converting low-resolution videos to visually appealing high-resolution ones. It may lead to temporal inconsistency due to the drastically changing outputs. In this paper, we propose cascaded sliding-window-based relativistic GAN (Generative Adversarial Network) fusion for perceptual and consistent video super-resolution (PC-VSR). Firstly, cascaded sliding-window-based relativistic GAN is designed to extract more useful information. It enlarges the temporal receptive field of sliding-window-based model in each step. It is able to enhance perceptual quality and compensate temporal consistency progressively and sufficiently. The trained separate refinement generator networks are fused into a final refinement generator. The final refinement generator can be calculated recursively at the testing stage. With our generator fusion, the parameter number is reduced and good quality is maintained. Extensive experimental results demonstrate that our approach outperforms state-of-the-art super-resolution methods in terms of perceptual quality. Our method also achieves good temporal consistency and per-pixel accuracy, compared with other perceptual approaches.
Document Type: conference object
Language: English
DOI: 10.1007/978-3-031-71253-1_17
Availability: https://inria.hal.science/hal-05142880
https://inria.hal.science/hal-05142880v1/document
https://inria.hal.science/hal-05142880v1/file/633143_1_En_17_Chapter.pdf
https://doi.org/10.1007/978-3-031-71253-1_17
Rights: http://creativecommons.org/licenses/by/
Accession Number: edsbas.7833912C
Database: BASE
Description
Abstract:Part 5: Perceptual Intelligence ; International audience ; Perceptual video super-resolution aims at converting low-resolution videos to visually appealing high-resolution ones. It may lead to temporal inconsistency due to the drastically changing outputs. In this paper, we propose cascaded sliding-window-based relativistic GAN (Generative Adversarial Network) fusion for perceptual and consistent video super-resolution (PC-VSR). Firstly, cascaded sliding-window-based relativistic GAN is designed to extract more useful information. It enlarges the temporal receptive field of sliding-window-based model in each step. It is able to enhance perceptual quality and compensate temporal consistency progressively and sufficiently. The trained separate refinement generator networks are fused into a final refinement generator. The final refinement generator can be calculated recursively at the testing stage. With our generator fusion, the parameter number is reduced and good quality is maintained. Extensive experimental results demonstrate that our approach outperforms state-of-the-art super-resolution methods in terms of perceptual quality. Our method also achieves good temporal consistency and per-pixel accuracy, compared with other perceptual approaches.
DOI:10.1007/978-3-031-71253-1_17