Presentation | 2021-11-25 グラフ学習に基づく光線空間上の雑音除去の検討 Rino Yoshida, Kazuya Kodama, Huy Vu, Gene Cheung, Takayuki Hamamoto, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | A light field (LF) image is composed of multi-view images acquired by slightly offset viewpoints. We propose a novel method to denoise noise-corrupted LF images based on graph learning. In comparison with the conventional deep learning for image restoration, our interpretable method combined with graph signal processing (GSP) needs to learn only far fewer parameters suitable for limited datasets. Given multi-view images acquired by dense viewpoints simply, neighboring pixels in the corresponding 4D LF image are connected to each other for GSP using graph total variation (GTV) as signal prior. In order to construct the graph for LF image denoising, we train a convolutional neural net (CNN) determining its edge weights. Experimental results show that our proposal outperformed model-based and deep-learning-based implementations with respect to denoising performance and robustness. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | Light field / denoising / multi-view images / graph total variation / convolutional neural nets |
Paper # | CS2021-62,IE2021-21 |
Date of Issue | 2021-11-18 (CS, IE) |
Conference Information | |
Committee | IPSJ-AVM / CS / IE / ITE-BCT |
---|---|
Conference Date | 2021/11/25(1days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | Online |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | Image coding, Communications and streaming technologies, etc. |
Chair | 笠井 裕之(早大) / Jun Terada(NTT) / Kazuya Kodama(NII) / Kyoichi Saito(NHK) |
Vice Chair | / Daisuke Umehara(Kyoto Inst. of Tech.) / Hiroyuki Bandoh(NTT) / Toshihiko Yamazaki(Univ. of Tokyo) / Hidekazu Murata(Kyoto Univ.) / Hajime Saito(TV Tokyo) |
Secretary | (早大) / Daisuke Umehara(KDDI) / Hiroyuki Bandoh(NTT) / Toshihiko Yamazaki(NICT) / Hidekazu Murata(NTT) / Hajime Saito(KDDI Research) |
Assistant | / Takahiro Yamaura(Toshiba) / Yuta Ida(Yamaguchi Univ.) / Shunsuke Iwamura(NHK) / Shinobu Kudo(NTT) / Tatsuhiko Itokawa(Mitsubishi Electric) / Takumi Matsumoto(Furukawa Electric) / Takeshi Maruyama(Furukawa Electric) / Yoshie Enoki(TBS) |
Paper Information | |
Registration To | The Special Interest Groups of Audio Visual and Multimedia Information Processing / Technical Committee on Communication Systems / Technical Committee on Image Engineering / Technical Group on Broadcasting and Communication Technologies |
---|---|
Language | JPN-ONLY |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | |
Sub Title (in English) | |
Keyword(1) | Light field |
Keyword(2) | denoising |
Keyword(3) | multi-view images |
Keyword(4) | graph total variation |
Keyword(5) | convolutional neural nets |
1st Author's Name | Rino Yoshida |
1st Author's Affiliation | Tokyo University of Science(TUS) |
2nd Author's Name | Kazuya Kodama |
2nd Author's Affiliation | National Institute of Informatics(NII) |
3rd Author's Name | Huy Vu |
3rd Author's Affiliation | York University(York Univ.) |
4th Author's Name | Gene Cheung |
4th Author's Affiliation | York University(York Univ.) |
5th Author's Name | Takayuki Hamamoto |
5th Author's Affiliation | Tokyo University of Science(TUS) |
Date | 2021-11-25 |
Paper # | CS2021-62,IE2021-21 |
Volume (vol) | vol.121 |
Number (no) | CS-268,IE-269 |
Page | pp.pp.13-18(CS), pp.13-18(IE), |
#Pages | 6 |
Date of Issue | 2021-11-18 (CS, IE) |