Ultra performance video codec
based on machine learning
SPIRIT DSP’S INNOVATIVE VIDEO CODEC
SPIRIT DSP has developed a fundamentally new concept of video coding (codec) that is based on neural networks, which increases video’s compression ratio by about 3 times, compared to traditional video codecs, by reducing statistical, psychophysical and applied redundancy.
SOLUTION OF THE YEAR 2020
Innovation and Excellence
Award by CorporateLiveWire
SPIRIT DSP’s innovative video codec:
pre-processes a video stream or video file;
transforms a video into a sketch form;
adapts the frame rate to the dynamics of the objects in a scene;
performs frame-by-frame entropy coding;
decodes sketch frames and restores the original frame rate;
restores video stream using a trained neural network;
provides an optional search for an object of interest, faces detection, stylistic coloring, etc.
Video content is projected to occupy 82% of the IP networks by 2021 (Cisco’s forecast).
The global video compression market is expected to reach $2.19 billion by 2023.
The urgent need to increase video compression efficiency is evident from the active
participation of the global technology leaders in the development of new video codecs. Some of
these players include ITU, Fraunhofer Society, MPEG (ISO), Alliance of Open Media (Apple,
Amazon, Google, Netflix, Facebook, Cisco, Intel, Microsoft, etc.). Examples of the new video
Н.266 (FVC), VP10, JVET, AV1 and others.
But all these developments are based on more and more complex coding algorithms. This traditional approach to coding, which has enabled developers to improve the characteristics of video codecs (by increasing the compression ratio, while maintaining the video’s original quality) on average by 50% in every 10 years, has now practically reached its fundamental limit.
Unlike the traditional approaches, the new method, patented by SPIRIT DSP, offers a video codec design paradigm shift by using deep learning and neural networks.
The advantages of SPIRIT DSP’s innovation:
significant reduction in costs of video data transmissions;
video streaming via narrowband and mobile communication channels.
SPIRIT DSP’s competitive positioning vs. some popular video codecs:
The author of this innovative concept is Dr. Vladimir Sviridenko, the CTO of SPIRIT DSP, whose distinguished 40 years of hands-on experience in data compression, includes, amongst others, the authorship of three monographs and over 150 articles, as well as the ownership of 45 patents. The latter include his patents on data compression, titled, “The Method and Device for Highly Efficiently Compressing Large Scale Multimedia Information According to the Criteria for Its Value for Storage in Data Storage Systems.” (officially registered under No. 2654126 in Russia on May 16, 2018).
The backbone of the team comprises specialists in image processing, computer vision and deep learning.
DSP is developing and licensing software products to the global market
for more than 25 years. SPIRIT communication software products are
licensed to/built into popular products from more than 250 global
technology leaders including Apple, Avaya, Adobe, AT&T, Blizzard,
BT, China Mobile, Ericsson, HP, HTC, Huawei, Korea Telecom, LG,
Microsoft, Oracle, Samsung, Skype, Toshiba, Viber, ZTE, etc. More than 1
billion people in more than 100 countries use SPIRIT software.
VideoMost® is a software videoconferencing server with PC and mobile clients that supports all the popular communication protocols and international standards, including H.323, H.239, SIP, BFCP, XMPP, H.264, G.7xx, WebRTC, and delivers interoperability with the legacy videoconferencing hardware. VideoMost works in popular browsers (Internet Explorer, Firefox, Chrome, Safari) and is available on both iOS and Android mobile platforms. It also provides a full range of tools for teamwork, including mobile messaging, document sharing and polling.