Making Film Grain Great Again: Introducing AV1-Compatible Film Grain Modeling for Existing HEVC-Based Video Codecs

Recently, film grain applications started to gain a lot of popularity due to an increasing demand for natural visual appearance, in the light of a dramatically increasing amount of artificially generated content. Film grain is spatially random in nature, similarly to noise, while its physical size can vary as well. In addition, film grain is independent in a temporal domain, thereby making it inherently difficult to compress for conventional video encoders. — Therefore, in order to improve a video coding gain it is desirable to remove the film grain prior to encoding, and then to add it back after decoding and prior to display as a post-processing step, thereby leading to significant bitrate savings. This approach was successfully incorporated by Alliance for Open Media (AOM) within its proprietary AV1 coding scheme as a normative process, starting from its 1^st version (AV1 v1.0) in 2018. — In this work, the AV1-compatible film grain modeling for the H.265/MPEG-HEVC based video codecs has been carried out to efficiently utilize the existing AV1 film grain post-processing support. This is done by providing the estimated film grain parameters within the ITU-T Recommendation T.35 (ITU-T T.35) Supplemental Enhancement Information (SEI) message. According to the extensive experimental results conducted on popular cinematic content, very significant bitrate savings are achieved for substantially the same subjective video presentation quality.

Published

2023-10

Content type

Original Research

Keywords

film grain, video compression, coding efficiency, video coding gain, HEVC, AOM, AV1, H.265, visual quality, subjective quality assessment

DOI

10.5594/M002021

ISBN

978-1-61482-964-5