Background

A method of generating interpolated video frames which allows content providers to reduce bandwidth of digital video delivery by removing video frames before sending content to a receiving device, reducing bandwidth needed for delivery or storage of video content by up to 50% with negligible perceptible quality loss to the viewer, achieving PSNR parity with industry standard MPEG‑based compression technologies (such as h.264/5 and V9).

Technology Overview

Using trained AI, the algorithm chooses anchor frames which will best retain the necessary image data to reconstruct the full video file after delivery, the remaining less relevant frames can then be removed and the anchor frames then compressed using the content owner’s preferred video compression technique (which may be open source, proprietary, or even encrypted).

The compressed video is then sent to the device which will store and/or reconstruct the video file. When called by the user, the video is decompressed (again using open source, proprietary, or encrypted decompression techniques determined by the content owner), and then the generative algorithm regenerates the non‑anchor frames that were removed prior to compression using AI‑based interpolation, fully restoring the video at optimal quality.

Benefits

The quality of the output is tunable, therefore the content owner can make a determination of what the quality of the decompressed/regenerated video will be based on factors relevant to their business model or user experience (such as: available bandwidth, remaining data available to the user, screen size/quality of the viewing device, tiered content quality paid by the user, computing power of the displaying device).

The regenerative video algorithm can be attached to any compression algorithm ‑ because the anchor frames are determined and superfluous frames are removed before compression and then regenerated after compression, this technology is agnostic as to compression algorithm and/or encryption utilized by the content owner.

Speed and quality are primarily a tradeoff of the power of the device; however, the algorithm has been crafted to take advantage of AI‑specific hardware and could be utilized on special‑purpose hardware or programmable cores.

Applications

Video streaming services
Video content storage
Video delivery providers (i.e. over the top content providers)
Cellular companies utilizing dynamic video compression to control network traffic

Opportunity

Looking for licensees and development partners, ideally those in a position to create a partnership across the video creation/storage/delivery/display industry.

Generative Digital Video Compression

Patents

IP Status

Seeking