BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Los_Angeles
X-LIC-LOCATION:America/Los_Angeles
BEGIN:DAYLIGHT
TZOFFSETFROM:-0800
TZOFFSETTO:-0700
TZNAME:PDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0700
TZOFFSETTO:-0800
TZNAME:PST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260402T024534Z
LOCATION:3001\, Level 3
DTSTART;TZID=America/Los_Angeles:20250624T114500
DTEND;TZID=America/Los_Angeles:20250624T120000
UID:dac_DAC 2025_sess102_RESEARCH313@linklings.com
SUMMARY:DCDiff: : Enhancing JPEG Compression via Diffusion-based DC Coeffi
 cients Estimation
DESCRIPTION:ZIYUAN ZHANG and Han Qiu (Tsinghua University), Tianwei Zhang 
 (Nanyang Technological University), Bin Chen (Harbin Institute of Technolo
 gy), and Chao Zhang (Tsinghua University)\n\nJPEG is the most widely-used 
 image compression method on low-cost cameras which cannot support learning
 -based compressors. One promising approach to enhance JPEG aims to drop DC
  coefficients at the cameras' ends (without extra computation) and reconst
 ruct those DC coefficients after receiving them. They all face the challen
 ge that their DC reconstruction relies on a statistical property, which wi
 ll cause deviation-introduced errors and propagate. In this paper, we prop
 ose DCDiff, a novel end-to-end DC estimation method to tackle the above ch
 allenge. Instead of using statistical methods to recover DC coefficients a
 nd then fix errors, we directly leverage a generative model to estimate DC
  coefficients in an end-to-end manner. In the meantime, we generate masks 
 to correct certain image locations that do not satisfy the statistical dis
 tribution to suppress error propagation. Extensive experiments show that D
 CDiff not only outperforms all baselines on compression performance but al
 so introduces a tiny impact on downstream tasks and is fully compatible wi
 th 2 typical low-cost processors with JPEG support.\n\nTopics: AI\n\nTrack
 s: AI1: AI/ML Algorithms\n\nSession Chairs: Chenhui Deng (Nvidia) and Liu 
 Cheng (Chinese Academy of Sciences, University of Chinese Academy of Scien
 ces)\n\n
END:VEVENT
END:VCALENDAR