ENTED: Enhanced Neural Texture Extraction and Distribution for Reference-based Blind Face Restoration

We present ENTED, a new framework for blind face restoration that aims to restore high-quality and realistic portrait images. Our method involves repairing a single degraded input image using a high-quality reference image. We utilize a texture extraction and distribution framework to transfer high-...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Lau, Yuen-Fui, Zhang, Tianjia, Rao, Zhefan, Chen, Qifeng
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page
container_issue
container_start_page
container_title
container_volume
creator Lau, Yuen-Fui
Zhang, Tianjia
Rao, Zhefan
Chen, Qifeng
description We present ENTED, a new framework for blind face restoration that aims to restore high-quality and realistic portrait images. Our method involves repairing a single degraded input image using a high-quality reference image. We utilize a texture extraction and distribution framework to transfer high-quality texture features between the degraded input and reference image. However, the StyleGAN-like architecture in our framework requires high-quality latent codes to generate realistic images. The latent code extracted from the degraded input image often contains corrupted features, making it difficult to align the semantic information from the input with the high-quality textures from the reference. To overcome this challenge, we employ two special techniques. The first technique, inspired by vector quantization, replaces corrupted semantic features with high-quality code words. The second technique generates style codes that carry photorealistic texture information from a more informative latent space developed using the high-quality features in the reference image's manifold. Extensive experiments conducted on synthetic and real-world datasets demonstrate that our method produces results with more realistic contextual details and outperforms state-of-the-art methods. A thorough ablation study confirms the effectiveness of each proposed module.
doi_str_mv 10.48550/arxiv.2401.06978
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_2401_06978</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2401_06978</sourcerecordid><originalsourceid>FETCH-LOGICAL-a678-387167bee12c034cf36f60263ab9edcce245f45ae98bacfbee5f293ecbf82b813</originalsourceid><addsrcrecordid>eNotj81OhDAUhbtxYUYfwJV9AbA_UMrsdIZRk8mYGPbkttxmmiCYAgbf3g7O6ibnfuckHyEPnKWZznP2BGHxP6nIGE-ZKgt9S7A61dV-S6v-DL3Flp5wDtDRGpdpDkirZQpgJz_0FPqW7v04BW_mNXBDoJ_oMGBsJgbGWH_pfMQOYDG-xmkIcEHvyI2DbsT7692Q-lDVu7fk-PH6vns-JqAKnUhdcFUYRC4sk5l1UjnFhJJgSmytRZHlLssBS23AugjmTpQSrXFaGM3lhjz-z66ezXfwXxB-m4tvs_rKP1_cUeE</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>ENTED: Enhanced Neural Texture Extraction and Distribution for Reference-based Blind Face Restoration</title><source>arXiv.org</source><creator>Lau, Yuen-Fui ; Zhang, Tianjia ; Rao, Zhefan ; Chen, Qifeng</creator><creatorcontrib>Lau, Yuen-Fui ; Zhang, Tianjia ; Rao, Zhefan ; Chen, Qifeng</creatorcontrib><description>We present ENTED, a new framework for blind face restoration that aims to restore high-quality and realistic portrait images. Our method involves repairing a single degraded input image using a high-quality reference image. We utilize a texture extraction and distribution framework to transfer high-quality texture features between the degraded input and reference image. However, the StyleGAN-like architecture in our framework requires high-quality latent codes to generate realistic images. The latent code extracted from the degraded input image often contains corrupted features, making it difficult to align the semantic information from the input with the high-quality textures from the reference. To overcome this challenge, we employ two special techniques. The first technique, inspired by vector quantization, replaces corrupted semantic features with high-quality code words. The second technique generates style codes that carry photorealistic texture information from a more informative latent space developed using the high-quality features in the reference image's manifold. Extensive experiments conducted on synthetic and real-world datasets demonstrate that our method produces results with more realistic contextual details and outperforms state-of-the-art methods. A thorough ablation study confirms the effectiveness of each proposed module.</description><identifier>DOI: 10.48550/arxiv.2401.06978</identifier><language>eng</language><subject>Computer Science - Computer Vision and Pattern Recognition</subject><creationdate>2024-01</creationdate><rights>http://arxiv.org/licenses/nonexclusive-distrib/1.0</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,776,881</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/2401.06978$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.2401.06978$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Lau, Yuen-Fui</creatorcontrib><creatorcontrib>Zhang, Tianjia</creatorcontrib><creatorcontrib>Rao, Zhefan</creatorcontrib><creatorcontrib>Chen, Qifeng</creatorcontrib><title>ENTED: Enhanced Neural Texture Extraction and Distribution for Reference-based Blind Face Restoration</title><description>We present ENTED, a new framework for blind face restoration that aims to restore high-quality and realistic portrait images. Our method involves repairing a single degraded input image using a high-quality reference image. We utilize a texture extraction and distribution framework to transfer high-quality texture features between the degraded input and reference image. However, the StyleGAN-like architecture in our framework requires high-quality latent codes to generate realistic images. The latent code extracted from the degraded input image often contains corrupted features, making it difficult to align the semantic information from the input with the high-quality textures from the reference. To overcome this challenge, we employ two special techniques. The first technique, inspired by vector quantization, replaces corrupted semantic features with high-quality code words. The second technique generates style codes that carry photorealistic texture information from a more informative latent space developed using the high-quality features in the reference image's manifold. Extensive experiments conducted on synthetic and real-world datasets demonstrate that our method produces results with more realistic contextual details and outperforms state-of-the-art methods. A thorough ablation study confirms the effectiveness of each proposed module.</description><subject>Computer Science - Computer Vision and Pattern Recognition</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2024</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotj81OhDAUhbtxYUYfwJV9AbA_UMrsdIZRk8mYGPbkttxmmiCYAgbf3g7O6ibnfuckHyEPnKWZznP2BGHxP6nIGE-ZKgt9S7A61dV-S6v-DL3Flp5wDtDRGpdpDkirZQpgJz_0FPqW7v04BW_mNXBDoJ_oMGBsJgbGWH_pfMQOYDG-xmkIcEHvyI2DbsT7692Q-lDVu7fk-PH6vns-JqAKnUhdcFUYRC4sk5l1UjnFhJJgSmytRZHlLssBS23AugjmTpQSrXFaGM3lhjz-z66ezXfwXxB-m4tvs_rKP1_cUeE</recordid><startdate>20240112</startdate><enddate>20240112</enddate><creator>Lau, Yuen-Fui</creator><creator>Zhang, Tianjia</creator><creator>Rao, Zhefan</creator><creator>Chen, Qifeng</creator><scope>AKY</scope><scope>GOX</scope></search><sort><creationdate>20240112</creationdate><title>ENTED: Enhanced Neural Texture Extraction and Distribution for Reference-based Blind Face Restoration</title><author>Lau, Yuen-Fui ; Zhang, Tianjia ; Rao, Zhefan ; Chen, Qifeng</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a678-387167bee12c034cf36f60263ab9edcce245f45ae98bacfbee5f293ecbf82b813</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2024</creationdate><topic>Computer Science - Computer Vision and Pattern Recognition</topic><toplevel>online_resources</toplevel><creatorcontrib>Lau, Yuen-Fui</creatorcontrib><creatorcontrib>Zhang, Tianjia</creatorcontrib><creatorcontrib>Rao, Zhefan</creatorcontrib><creatorcontrib>Chen, Qifeng</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Lau, Yuen-Fui</au><au>Zhang, Tianjia</au><au>Rao, Zhefan</au><au>Chen, Qifeng</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>ENTED: Enhanced Neural Texture Extraction and Distribution for Reference-based Blind Face Restoration</atitle><date>2024-01-12</date><risdate>2024</risdate><abstract>We present ENTED, a new framework for blind face restoration that aims to restore high-quality and realistic portrait images. Our method involves repairing a single degraded input image using a high-quality reference image. We utilize a texture extraction and distribution framework to transfer high-quality texture features between the degraded input and reference image. However, the StyleGAN-like architecture in our framework requires high-quality latent codes to generate realistic images. The latent code extracted from the degraded input image often contains corrupted features, making it difficult to align the semantic information from the input with the high-quality textures from the reference. To overcome this challenge, we employ two special techniques. The first technique, inspired by vector quantization, replaces corrupted semantic features with high-quality code words. The second technique generates style codes that carry photorealistic texture information from a more informative latent space developed using the high-quality features in the reference image's manifold. Extensive experiments conducted on synthetic and real-world datasets demonstrate that our method produces results with more realistic contextual details and outperforms state-of-the-art methods. A thorough ablation study confirms the effectiveness of each proposed module.</abstract><doi>10.48550/arxiv.2401.06978</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.2401.06978
ispartof
issn
language eng
recordid cdi_arxiv_primary_2401_06978
source arXiv.org
subjects Computer Science - Computer Vision and Pattern Recognition
title ENTED: Enhanced Neural Texture Extraction and Distribution for Reference-based Blind Face Restoration
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-02-03T22%3A37%3A36IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-arxiv_GOX&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=ENTED:%20Enhanced%20Neural%20Texture%20Extraction%20and%20Distribution%20for%20Reference-based%20Blind%20Face%20Restoration&rft.au=Lau,%20Yuen-Fui&rft.date=2024-01-12&rft_id=info:doi/10.48550/arxiv.2401.06978&rft_dat=%3Carxiv_GOX%3E2401_06978%3C/arxiv_GOX%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_id=info:pmid/&rfr_iscdi=true