iSIZE - BitClear
Category Streaming - Video processing
User-generated content (UGC) that is viewed by millions of people has been re-shared and re-uploaded numerous times, which tends to make it very degraded due to multiple transcoding iterations happening on non-optimized platforms, e.g., mobile phones transcoding for bandwidth-constrained uploads. For social media or UGC distribution and streaming companies, presenting video content in as high quality as possible is vital to keeping audiences engaged and maximize their video asset.
An AI-based video processing technology, BitClear makes unwatchable videos watchable by removing compression artifacts (blurring, blocking, etc.) from user-generated, or heavily compressed content. Video content exposed to multiple transcoding iterations can be revived to the maximum possible quality without affecting the original artistic intent of its creators. The process also allows for video upscaling, all with as little as 30ms processing latency on GPUs or high-performance CPUs.
Leveraging bespoke neural network designs that optimally learns to disentangle the noise from the data manifolds, BitClear removes compression noise and retains or recovers the original content features. While it learns the noise signatures of the various encoding standards, BitClear does not need to know the history of the specific asset: it can process any highly compressed content and produce a higher-quality output that improves the value of the asset.
Unlike previous noise removal technologies aimed at specific noise patterns, like film grain, blur, or interlacing artifacts, BitClear specifically targets the compression artifacts of typical MPEG or AOMedia encoders, focusing on highly compressed or highly distorted content.
Its bespoke neural network architecture makes it readily scalable, up to huge volumes of parallel processing. On an Intel Xeon CPU, the fastest BitClear models achieve 30 frames per second when creating 1080p output content. In addition, on NVIDIA RTX, T4 or V100 GPUs, BitClear can operate over multiple input video assets in real time (at 25/30 fps). The capability to scale up to high volumes of content without user intervention makes BitClear ideal for cloud as well as on-premise deployment on CPUs, GPUs or custom hardware that supports neural network inference. Its AI nature and implementation efficiency means it can be deployed at scale without the need for human inspection or tuning.
BitClear’s underlying AI principles can be applied to other applications that require the separation of image from noise and distortion. For example, restoration of archive content, with the BitClear denoising engine repurposed for this target. It can also be applicable for cases where delivery bandwidth is limited by infrastructure and BitClear can denoise the content prior to display.
Examples include, iin broadcast contribution feeds over mobile streaming, or for gaming, live sports, entertainment and virtual reality, where BitClear can run in real-time on a mobile NPU on the device, IoT applications like security that start with very high image quality, at 1080p or higher, but are severely constrained on the upstream bandwidth, where BitClear can run on the datacenter and denoise the received content prior to onwards processing. There is broad scope for real-time or near real-time video denoising across many industries.
