Nokia has officially released an open-source implementation of its Neural Network Post Filter (NNPF) technology under a BSD 3-Clause Clear License, marking the first time an AI-based video enhancement method has been integrated into international video coding standards. The open-source release is designed to improve compressed video quality during playback on consumer devices without requiring higher bitrates or replacing existing video codecs.
Rather than introducing a new compression format, NNPF operates after a video is decoded on a user’s device. It utilizes neural network processing to restore fine visual details, sharpen textures, smooth color gradients, and eliminate compression artifacts like color banding. The system relies on Supplemental Enhancement Information (SEI) metadata delivered alongside the video stream under the Versatile Supplemental Enhancement Information (VSEI) standard. This allows encoders to specify and fine-tune appropriate neural network models for specific video sequences, which compatible smart TVs, smartphones, and PCs can then execute in real time using local Neural Processing Units (NPUs) or GPUs.
Because NNPF works natively alongside established compression formats like High Efficiency Video Coding (HEVC) and Versatile Video Coding (VVC), content providers and streaming platforms can deliver perceptually higher-quality video without re-encoding existing libraries or increasing bandwidth demand. By open-sourcing the software implementation—which integrates VVC decoding with real-time AI inference via Intel’s OpenVINO framework—Nokia aims to accelerate industry adoption across hardware manufacturers, app developers, and streaming platforms. While widespread commercial availability will depend on ecosystem integration across chipmakers and playback applications, the open standard offers a scalable, royalty-accessible path toward AI-driven video optimization.
