Learning across views for stereo image completion
Stereo image completion (SIC) is to fill holes existing in a pair of stereo images. SIC is more complicated than single image repairing, which needs to complete the pair of images while keeping their stereoscopic consistency. In recent years, deep learning has been introduced into single image repai...
Main Authors: | , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
Wiley
2020-10-01
|
Series: | IET Computer Vision |
Subjects: | |
Online Access: | https://doi.org/10.1049/iet-cvi.2019.0775 |
_version_ | 1797684644119838720 |
---|---|
author | Wei Ma Mana Zheng Wenguang Ma Shibiao Xu Xiaopeng Zhang |
author_facet | Wei Ma Mana Zheng Wenguang Ma Shibiao Xu Xiaopeng Zhang |
author_sort | Wei Ma |
collection | DOAJ |
description | Stereo image completion (SIC) is to fill holes existing in a pair of stereo images. SIC is more complicated than single image repairing, which needs to complete the pair of images while keeping their stereoscopic consistency. In recent years, deep learning has been introduced into single image repairing but seldom used for SIC. The authors present a novel deep learning‐based approach for SIC. In their method, an X‐shaped fully convolutional network (called SICNet) is proposed and designed to complete stereo images, which is composed of two branches of convolutional neural network layers to encode the context of the left and right images separately, a fusion module for stereo‐interactive completion, and two branches of decoders to produce completed left and right images, respectively. In consideration of both inter‐view and intra‐view cues, they introduce auxiliary networks and define comprehensive losses to train SICNet to perform single‐view coherent and cross‐view consistent completion simultaneously. Extensive experiments are conducted to show the state‐of‐the‐art performances of the proposed approach and its key components. |
first_indexed | 2024-03-12T00:32:44Z |
format | Article |
id | doaj.art-f42e0b5f131e4bb7b95844213aa646d3 |
institution | Directory Open Access Journal |
issn | 1751-9632 1751-9640 |
language | English |
last_indexed | 2024-03-12T00:32:44Z |
publishDate | 2020-10-01 |
publisher | Wiley |
record_format | Article |
series | IET Computer Vision |
spelling | doaj.art-f42e0b5f131e4bb7b95844213aa646d32023-09-15T10:11:27ZengWileyIET Computer Vision1751-96321751-96402020-10-0114748249210.1049/iet-cvi.2019.0775Learning across views for stereo image completionWei Ma0Mana Zheng1Wenguang Ma2Shibiao Xu3Xiaopeng Zhang4Faculty of Information TechnologyBeijing University of TechnologyNo. 100 Pingleyuan Street, Chaoyang DistrictBeijingPeople's Republic of ChinaFaculty of Information TechnologyBeijing University of TechnologyNo. 100 Pingleyuan Street, Chaoyang DistrictBeijingPeople's Republic of ChinaFaculty of Information TechnologyBeijing University of TechnologyNo. 100 Pingleyuan Street, Chaoyang DistrictBeijingPeople's Republic of ChinaNational Laboratory of Pattern RecognitionInstitute of AutomationChinese Academy of SciencesBeijing100190People's Republic of ChinaNational Laboratory of Pattern RecognitionInstitute of AutomationChinese Academy of SciencesBeijing100190People's Republic of ChinaStereo image completion (SIC) is to fill holes existing in a pair of stereo images. SIC is more complicated than single image repairing, which needs to complete the pair of images while keeping their stereoscopic consistency. In recent years, deep learning has been introduced into single image repairing but seldom used for SIC. The authors present a novel deep learning‐based approach for SIC. In their method, an X‐shaped fully convolutional network (called SICNet) is proposed and designed to complete stereo images, which is composed of two branches of convolutional neural network layers to encode the context of the left and right images separately, a fusion module for stereo‐interactive completion, and two branches of decoders to produce completed left and right images, respectively. In consideration of both inter‐view and intra‐view cues, they introduce auxiliary networks and define comprehensive losses to train SICNet to perform single‐view coherent and cross‐view consistent completion simultaneously. Extensive experiments are conducted to show the state‐of‐the‐art performances of the proposed approach and its key components.https://doi.org/10.1049/iet-cvi.2019.0775convolutional neural network layersstereo‐interactive completioncross‐view consistent completionstereo image completionsingle image repairingdeep learning |
spellingShingle | Wei Ma Mana Zheng Wenguang Ma Shibiao Xu Xiaopeng Zhang Learning across views for stereo image completion IET Computer Vision convolutional neural network layers stereo‐interactive completion cross‐view consistent completion stereo image completion single image repairing deep learning |
title | Learning across views for stereo image completion |
title_full | Learning across views for stereo image completion |
title_fullStr | Learning across views for stereo image completion |
title_full_unstemmed | Learning across views for stereo image completion |
title_short | Learning across views for stereo image completion |
title_sort | learning across views for stereo image completion |
topic | convolutional neural network layers stereo‐interactive completion cross‐view consistent completion stereo image completion single image repairing deep learning |
url | https://doi.org/10.1049/iet-cvi.2019.0775 |
work_keys_str_mv | AT weima learningacrossviewsforstereoimagecompletion AT manazheng learningacrossviewsforstereoimagecompletion AT wenguangma learningacrossviewsforstereoimagecompletion AT shibiaoxu learningacrossviewsforstereoimagecompletion AT xiaopengzhang learningacrossviewsforstereoimagecompletion |