BEGIN:VCALENDAR VERSION:2.0 PRODID:-//swoogo.com//NONSGML kigkonsult.se iCalcreator 2.41.90// CALSCALE:GREGORIAN UID:32643531-3761-4339-b833-666365363839 BEGIN:VEVENT UID:d08b8d78371c5fe2be7542d802873ada40e4dc55@swoogo.com DTSTAMP:20260713T095507Z DESCRIPTION:Film grain\, integral to analogue cinema's aesthetic\, challeng es video distribution in low-bitrate scenarios\, often compromising its au thenticity due to compression constraints. AV1 film grain synthesis techno logy addresses this by separating and estimating film grain parameters dur ing encoding\, which are then used to reconstruct grain during decoding. H owever\, the choice of film grain block size impacts the quality of recons truction\, with larger blocks potentially introducing noticeable repetitiv e noise. We propose a novel perceptual quality assessment method to detect and penalize these repetitive patterns. By leveraging the human visual sy stem's masking effect\, our method normalizes decoded frames and assesses local cosine similarities with the film grain template to identify and qua ntify noise patterns. Implemented in the frequency domain\, it processes h igh-resolution videos efficiently (around 2 FPS for UHD frames). Subjectiv e experiments with expert viewers confirm high correlations (SRCC = 0.853\ , PLCC = 0.856) between our method and human judgments\, validating its ef fectiveness in preserving film grain authenticity. DTSTART:20241023T170000Z DTEND:20241023T173000Z LAST-MODIFIED:20260713T095507Z LOCATION:Session Room 1 SEQUENCE:0 STATUS:CONFIRMED SUMMARY:AV1 Repetitive Film Grain Pattern Detection and Quality Assessment TRANSP:OPAQUE X-ALT-DESC;FMTTYPE=text/html:

Film grain\, integral to analogue cinema's aesthetic\, challenges video distribution in low-bitrate scenarios\, often compromising its authenticity due to compression constraints. AV1 film gr ain synthesis technology addresses this by separating and estimating film grain parameters during encoding\, which are then used to reconstruct grai n during decoding. However\, the choice of film grain block size impacts t he quality of reconstruction\, with larger blocks potentially introducing noticeable repetitive noise. We propose a novel perceptual quality assessm ent method to detect and penalize these repetitive patterns. By leveraging the human visual system's masking effect\, our method normalizes decoded frames and assesses local cosine similarities with the film grain template to identify and quantify noise patterns. Implemented in the frequency dom ain\, it processes high-resolution videos efficiently (around 2 FPS for UH D frames). Subjective experiments with expert viewers confirm high correla tions (SRCC = 0.853\, PLCC = 0.856) between our method and human judgments \, validating its effectiveness in preserving film grain authenticity.

BEGIN:VALARM UID:39323330-3037-4434-b365-633566626666 ACTION:DISPLAY DESCRIPTION:Film grain\, integral to analogue cinema's aesthetic\, challeng es video distribution in low-bitrate scenarios\, often compromising its au thenticity due to compression constraints. AV1 film grain synthesis techno logy addresses this by separating and estimating film grain parameters dur ing encoding\, which are then used to reconstruct grain during decoding. H owever\, the choice of film grain block size impacts the quality of recons truction\, with larger blocks potentially introducing noticeable repetitiv e noise. We propose a novel perceptual quality assessment method to detect and penalize these repetitive patterns. By leveraging the human visual sy stem's masking effect\, our method normalizes decoded frames and assesses local cosine similarities with the film grain template to identify and qua ntify noise patterns. Implemented in the frequency domain\, it processes h igh-resolution videos efficiently (around 2 FPS for UHD frames). Subjectiv e experiments with expert viewers confirm high correlations (SRCC = 0.853\ , PLCC = 0.856) between our method and human judgments\, validating its ef fectiveness in preserving film grain authenticity. TRIGGER:-PT15M END:VALARM END:VEVENT END:VCALENDAR