BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Australia/Melbourne
X-LIC-LOCATION:Australia/Melbourne
BEGIN:DAYLIGHT
TZOFFSETFROM:+1000
TZOFFSETTO:+1100
TZNAME:AEDT
DTSTART:19721003T020000
RRULE:FREQ=YEARLY;BYMONTH=4;BYDAY=1SU
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:19721003T020000
TZOFFSETFROM:+1100
TZOFFSETTO:+1000
TZNAME:AEST
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260114T163643Z
LOCATION:Meeting Room C4.11\, Level 4 (Convention Centre)
DTSTART;TZID=Australia/Melbourne:20231214T165300
DTEND;TZID=Australia/Melbourne:20231214T170300
UID:siggraphasia_SIGGRAPH Asia 2023_sess133_papers_812@linklings.com
SUMMARY:Diffusion-based Holistic Texture Rectification and Synthesis
DESCRIPTION:Guoqing Hao (University of Tsukuba, National Institute of Adva
 nced Industrial Science and Technology); Satoshi Iizuka (University of Tsu
 kuba); Kensho Hara (National Institute of Advanced Industrial Science); Ed
 gar Simo-Serra (Waseda University); Hirokatsu Kataoka (National Institute 
 of Advanced Industrial Science); and Kazuhiro Fukui (University of Tsukuba
 )\n\nWe present a novel framework for rectifying occlusions and distortion
 s in degraded texture samples from natural images. Traditional texture syn
 thesis approaches focus on generating textures from pristine samples, whic
 h necessitate meticulous preparation by humans and are often unattainable 
 in most natural images. These challenges stem from the frequent occlusions
  and distortions of texture samples in natural images due to obstructions 
 and variations in object surface geometry. To address these issues, we pro
 pose a framework that synthesizes holistic textures from degraded samples 
 in natural images, extending the applicability of exemplar-based texture s
 ynthesis techniques. Our framework utilizes a conditional Latent Diffusion
  Model (LDM) with a novel occlusion-aware latent transformer. This latent 
 transformer not only effectively encodes texture features from partially-o
 bserved samples necessary for the generation process of the LDM, but also 
 explicitly captures long-range dependencies in samples with large occlusio
 ns. To train our model, we introduce a method for generating synthetic dat
 a by applying geometric transformations and free-form mask generation to c
 lean textures. Experimental results demonstrate that our framework signifi
 cantly outperforms existing methods both quantitatively and quantitatively
 . Furthermore, we conduct comprehensive ablation studies to validate the d
 ifferent components of our proposed framework. Results are corroborated by
  a perceptual user study which highlights the efficiency of our proposed a
 pproach.\n\nRegistration Category: Full Access\n\nSession Chair: Anton Kap
 lanyan (Intel)\n\n
URL:https://asia.siggraph.org/2023/full-program?id=papers_812&sess=sess133
END:VEVENT
END:VCALENDAR