The iSIZE team was at this year’s Alliance for Open Media Research Symposium, in Redwood City from 7-8 June. CTO Dr Yiannis Andreopoulos gave a presentation on Day 1, speaking about neural pre and post-processing for video encoding with a range of codecs such as AVC, VP9 and AV1.
With more consumers and corporates streaming an ever-growing amount of video, quality remains a key differentiator. Yiannis highlighted how iSIZE BitSave preprocessing and BitClear post-processing can optimize the existing delivery pipeline to improve encoding quality as measured by standard perceptual quality metrics. During his presentation, Yiannis also explained how the iSIZE AI engine requires no input form the encoder, instead learning to distinguish perceptually unnoticeable details in content in an autonomous way. The result is a 12-15% increase in compression efficacy across AVC, VP9 and AV1. BitSave and BitClear can be used individually or in tandem over existing AVC and AOMedia encoders. When deployed together, BitSave and BitClear can deliver a -20% BD-rate savings over the corresponding encoder.
Optimizing video is not just about a better viewer experience, it also creates a significant reduction in the environmental impact of streaming, something that the industry as a whole is under pressure to address. Applying iSIZE’s AI-based pre-processing before encoding makes ingesting encoded content much easier and more efficient, cutting the overall carbon footprint.
The annual AOMedia Research Symposium is aimed at driving innovation in the digital media industry and fostering collaboration with the global academic community.
Click here for the full presentation.