BitSave

BitSave Tech Summary

BitSave Tech Summary

Over the last three decades, advances in video coding algorithms have led to ever improving compression rates. With video occupying more than 80% of all IP traffic by 2022, the end of the quest for bitrate saving is not in sight: every network upgrade is quickly diminished by the increasing demand for content. A multitude of different video coding standards are available to tackle this issue, with every codec generation aiming at 30-50% increase in compression efficiency at the expense of increasing complexity. The latter aspect means that new standards are being adopted very slowly: for instance, while MPEG HEVC was standardized in 2013, its high licensing fees and complexity (especially for HD and 4K content) have not yet allowed for wide adoption.

Where does iSIZE fit in? iSIZE is the first company to offer proprietary machine learning solutions for substantial bitrate or quality gains in video compression. Beyond its performance, what makes our solution stand out is that it is compatible with any existing video coding infrastructure. Therefore, it can:

• boost the compression efficiency of any video codec;
• run on client devices with minimal or no additional overhead;
• offer significant computational and energy efficiency for video encoding on resource-constrained devices (drones, action-cams, smartphones, etc.).

These advantages allow our clients to seamlessly integrate our solution and benefit from bitrate saving and quality improvement for their video delivery services.

Our Core Innovation

Future coding standards, like the ongoing VCEG/MPEG JVET and the AOMedia AV2 standardizations will create the next generation codecs that will replace HEVC and VP9/AV1. Such efforts undertake a lengthy development process that will typically culminate in 30%-40% bitrate saving for the same visual quality. However, the expected timeline for delivery of the first working codecs for MPEG’s JVET and AOMedia’s AV2 standardizations are scheduled for after 2023 – at the same time, the HEVC standard (finalized in 2013) has still not reached large rollout to date. On the other hand, current machine learning solutions like Magic Pony (Twitter) and Wave One offer disruptive performance for still-image coding; however, such solutions face substantial barriers when moved to video due to the unresolved challenge of incorporating temporal prediction and their deployment complexity.

Our bitrate saving and quality improvements are achieved by incorporating iSize’s proprietary deep perceptual optimisation and precoding technologies as a preprocessing stage of a standard codec pipeline. The entire process is shown in Figure 1. Instead of abandoning the existing codec pipeline (as proposed by deep autoencoding solutions, such as Magic Pony/Twitter and Wave One), iSIZE’s encoder-side solution preprocesses the input content with a custom-designed deep neural network solution. No changes are required in the encoding, stream packaging, streaming and decoding sides to obtain the final result. Our IP offers up to 40% bitrate saving over a wide range of encoding standards and encoding recipes, or commensurate quality improvement (e.g., 6 to 10 point increase in VMAF or similar high-level perceptual quality metrics). These gains can be used to also reduce encoding complexity, thereby saving datacenter processing power and energy consumption.

Unlike other machine learning efforts in this space, our solution is deployable today, and can be used on top of any standards-compliant or proprietary video codec architecture with minimal increase in complexity.

Product Specs