The paper proposes a novel frame-wise view synthesis method based on convolutional neural networks (CNNs) for wide-baseline light field (LF) camera arrays. A novel neural network architecture that follows a multi-resolution processing paradigm is employed to synthesize an entire view. A novel loss function formulation based on the structural similarity index (SSIM) is proposed. A wide-baseline LF image dataset is generated and employed to train the proposed deep model. The proposed method synthesizes each subaperture image (SAI) from a LF image based on corresponding SAIs from two reference LF images. Experimental results show that the proposed method yields promising results with an average PSNR and SSIM of 34.71 dB and 0.9673 respectively for wide baselines.
Schiopu, I, Rondao Alface, P & Munteanu, A 2019, Frame-wise CNN-based View Synthesis for Light field Camera Arrays. in International Conference on 3D Immersion. 2019 edn, 58, 2019 International Conference on 3D Immersion, IC3D 2019 - Proceedings, vol. 2019-January, IEEE, pp. 1, International Conference on 3D Immersion, Brussels, Belgium, 11/12/19. https://doi.org/10.1109/IC3D48390.2019.8975901
Schiopu, I., Rondao Alface, P., & Munteanu, A. (2019). Frame-wise CNN-based View Synthesis for Light field Camera Arrays. In International Conference on 3D Immersion (2019 ed., pp. 1). Article 58 (2019 International Conference on 3D Immersion, IC3D 2019 - Proceedings; Vol. 2019-January). IEEE. https://doi.org/10.1109/IC3D48390.2019.8975901
@inproceedings{73ad385d57f54e2cbfceb01e588a6813,
title = "Frame-wise CNN-based View Synthesis for Light field Camera Arrays",
abstract = "The paper proposes a novel frame-wise view synthesis method based on convolutional neural networks (CNNs) for wide-baseline light field (LF) camera arrays. A novel neural network architecture that follows a multi-resolution processing paradigm is employed to synthesize an entire view. A novel loss function formulation based on the structural similarity index (SSIM) is proposed. A wide-baseline LF image dataset is generated and employed to train the proposed deep model. The proposed method synthesizes each subaperture image (SAI) from a LF image based on corresponding SAIs from two reference LF images. Experimental results show that the proposed method yields promising results with an average PSNR and SSIM of 34.71 dB and 0.9673 respectively for wide baselines.",
keywords = "Machine Learning, View synthesis, Deep Learning",
author = "Ionut Schiopu and {Rondao Alface}, Patrice and Adrian Munteanu",
year = "2019",
month = dec,
day = "11",
doi = "10.1109/IC3D48390.2019.8975901",
language = "English",
series = "2019 International Conference on 3D Immersion, IC3D 2019 - Proceedings",
publisher = "IEEE",
pages = "1",
booktitle = "International Conference on 3D Immersion",
edition = "2019",
note = "International Conference on 3D Immersion, IC3D 2019 ; Conference date: 11-12-2019 Through 11-12-2019",
url = "https://www.stereopsia.com/international-conference-3d-immersion-ic3d",
}