Biological Image and Video Compression

生物图像和视频压缩

• 批准号: 105768
• 金额: $ 63.5万
• 依托单位: DEEP RENDER LTD
• 依托单位国家: 英国
• 项目类别: Study
• 财政年份: 2019
• 资助国家: 英国
• 起止时间: 2019 至 无数据
• 项目状态: 已结题
• 来源: https://gtr.ukri.org/projects?ref=105768
• 关键词: Biological; Image; Video; Compression

Deep Render Ltd is an Imperial College spin-out of the Department of Computing. We are a London based AI start-up that is developing the next generation of media compression algorithms. Our proprietary and patented technology is at the forefront of machine learning research. Furthermore, our Biological Image Compression results are state-of-the-art, already providing a 50% efficiency gain over the best previous compression standards.Our vision is to combine the fields of artificial intelligence, statistics and information theory to unlock the fundamental limits of image and video compression. The human eye is the best data compressor known to humanity -- with compression ratios at least 2,000 times better than everything developed to date. Biological Compression approximates the neurological processes of the human eye through a non-linear, learning-based approach, thereby creating a novel class of highly efficient compression algorithms. Moreover, while the traditional compression methods have hit peak-innovation with significantly declining performance gains, Biological Compression technology is only starting.Once Deep Render has completed the project, we will have an image compression codec at least 75% more efficient than the current state-of-the-art. Such an efficiency jump is the most significant inter-generation improvement the compression industry has ever seen. Further, we will have started to roll over these advances to video compression as well.Our target customers are the streaming industry (Netflix, YouTube, BBC, Sky), as well as the cloud-storage industry for images (Facebook, Instagram, Snapchat). Niche markets include VR/AR-streaming (Oculus), ultra-low latency live streaming (video calls, Skype), and all industries with large image/video storage needs such as the medical image processing market (NHS).Our value proposition is easy to understand. By making file sizes 75% smaller, we directly increase the bandwidth supply of the internet by a factor of 4, thus reducing data transport costs and operational overhead. Increasing the bandwidth supply by making file sizes smaller, is magnitudes more cost- and time-efficient than increasing the bandwidth supply through rewiring the globe with progressively more fibre-cables.Deep Render is going to help create a new age in which bandwidth constraints are a problem of the past. We are excitingly looking forward to seeing what amazing products people will build in such a world.

Deep Render Ltd是帝国理工学院计算机系的一个分支。我们是一家位于伦敦的人工智能初创公司，正在开发下一代媒体压缩算法。我们的专有和专利技术处于机器学习研究的最前沿。此外，我们的生物图像压缩结果是最先进的，已经提供了50%的效率增益比以前最好的压缩标准。我们的愿景是联合收割机结合人工智能，统计学和信息理论的领域，以解锁图像和视频压缩的基本限制。人眼是人类已知的最好的数据压缩器-压缩比至少比迄今为止开发的任何东西都好2,000倍。生物压缩通过一种非线性的、基于学习的方法来近似人眼的神经过程，从而创建了一类新型的高效压缩算法。此外，当传统的压缩方法已经达到了顶峰，但性能提升却在显著下降时，生物压缩技术才刚刚起步。一旦Deep Render完成项目，我们将拥有一个比当前最先进的图像压缩编解码器效率至少高75%的图像压缩编解码器。这样的效率跃升是压缩行业有史以来最显著的跨代改进。我们的目标客户是流媒体行业（Netflix、YouTube、BBC、Sky）以及图像云存储行业（Facebook、Instagram、Snapchat）。利基市场包括VR/AR流媒体（Oculus），超低延迟直播流媒体（视频通话，Skype）以及所有具有大型图像/视频存储需求的行业，如医疗图像处理市场（NHS）。我们的价值主张很容易理解。通过将文件大小缩小75%，我们直接将互联网的带宽供应增加了4倍，从而降低了数据传输成本和运营开销。通过减小文件大小来增加带宽供应，比通过用越来越多的光纤电缆重新布线地球仪来增加带宽供应更节省成本和时间。Deep Render将帮助创造一个新时代，在这个时代，带宽限制将成为过去的问题。我们非常期待看到人们在这样一个世界中会创造出什么样的惊人产品。