Single Image Reflection Removal Based on GAN with Gradient Constraint

Ryo Abiko, Masaaki Ikehara

    Research output: Chapter in Book/Report/Conference proceedingConference contribution

    Abstract

    When we take a picture through glass windows, the photographs are often degraded by undesired reflections. To separate reflection layer and background layer is an important problem for enhancing image quality. However, single-image reflection removal is a challenging process because of the ill-posed nature of the problem. In this paper, we propose a single-image reflection removal method based on generative adversarial network. Our network is an end-to-end trained network with four types of losses. It includes pixel loss, feature loss, adversarial loss and gradient constraint loss. We propose a novel gradient constraint loss in order to separate the background layer and the reflection layer clearly. Gradient constraint loss is applied in a gradient domain and it minimize the correlation between the background and reflection layer. Owing to the novel loss and our new synthetic dataset, our reflection removal method outperforms state-of-the-art methods in PSNR and SSIM, especially in real world images.

    Original languageEnglish
    Title of host publicationPattern Recognition - 5th Asian Conference, ACPR 2019, Revised Selected Papers
    EditorsShivakumara Palaiahnakote, Gabriella Sanniti di Baja, Liang Wang, Wei Qi Yan
    PublisherSpringer
    Pages609-624
    Number of pages16
    ISBN (Print)9783030414030
    DOIs
    Publication statusPublished - 2020 Jan 1
    Event5th Asian Conference on Pattern Recognition, ACPR 2019 - Auckland, New Zealand
    Duration: 2019 Nov 262019 Nov 29

    Publication series

    NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
    Volume12046 LNCS
    ISSN (Print)0302-9743
    ISSN (Electronic)1611-3349

    Conference

    Conference5th Asian Conference on Pattern Recognition, ACPR 2019
    CountryNew Zealand
    CityAuckland
    Period19/11/2619/11/29

    Keywords

    • Deep learning
    • Generative adversarial network
    • Image separation
    • Reflection removal

    ASJC Scopus subject areas

    • Theoretical Computer Science
    • Computer Science(all)

    Fingerprint Dive into the research topics of 'Single Image Reflection Removal Based on GAN with Gradient Constraint'. Together they form a unique fingerprint.

  • Cite this

    Abiko, R., & Ikehara, M. (2020). Single Image Reflection Removal Based on GAN with Gradient Constraint. In S. Palaiahnakote, G. Sanniti di Baja, L. Wang, & W. Q. Yan (Eds.), Pattern Recognition - 5th Asian Conference, ACPR 2019, Revised Selected Papers (pp. 609-624). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 12046 LNCS). Springer. https://doi.org/10.1007/978-3-030-41404-7_43