Recently, the 2022 MSU World Video Encoder Contest was officially released. NetEase Yunxin achieved impressive results for the first time in the competition. The self-developed NE265E encoder combined with smart code ultra-clear technology participated in an open identity and won many awards under the H.265 track. The index ranks first.
Competing for the first time, won the first place in 3 indicators
MSU Video Codecs Comparison 2022
MSU Video Codecs Comparison is an international competition sponsored by Moscow State University. It has been held for 17 consecutive sessions so far. It is known as the most authoritative and influential global top event in the field of video coding and compression.
In the FullHD competition, there are even more masters, almost covering many experts in the field of audio and video at home and abroad. The various video encoders participating in the competition are required to evaluate the encoding quality layer by layer through various objective quality indicators and subjective quality indicators under certain rules such as speed and bit rate, and give the final result of the competition.
In this competition, NetEase Yunxin NE265 encoder participated for the first time, and won the first place in 3 indicators and the second place in 8 indicators in the 15 indicators of the H.265 track as a public encoder.
Among them, in terms of the objective indicators of Y-VMAF (v0.6.1/v0.6.2/v0.6.3), NetEase Yunxinzhicode ultra-clear NE265E encoder has better performance than other encoders at all speeds of 1fps/5fps/30fps Dachang's participating encoder won the first place in the H.265 track. In terms of other objective quality indicators such as YUV-SSIM, YUV-MS-SSIM, YUV-PSNR, etc., at 5fps/30fps speed gear, NetEase Yunxin NE265 encoder won the second place in the H.265 track.
In high-level international top-level competitions, many major manufacturers competed on the same stage. NetEase Yunxin was able to participate in the competition with an open encoder and won the first place in multiple indicators of the segmented track, which shows the leading position of this smart code ultra-clear technology based on the NE265E encoder It has been recognized by the industry authority.
Netease Yunxin Zhicode ultra-clear NE265E encoder won the first place in the track at all speeds of 1fps/5fps/30fps
What can such awesome technology be used for?
MSU Video Codecs Comparison 2022
Some people may ask, this technology sounds awesome, but those indicators are not understood by the people who eat melons. What does it have to do with daily life?
There really are. As video becomes the media carrier of many contents, people have higher and higher requirements for video quality, from 1080P, 2K to 4K, and even 8K. However, if the viewing experience is better, the resolution size must be larger and larger. If it is not compressed by a video encoder, the cost of network bandwidth traffic will also increase.
Netease Yunxin Zhicode ultra-clear technology solves this problem, and can provide users with a higher quality video experience at the same bit rate.
In this way, for streaming media platforms, the cost of video storage, distribution, and dissemination can be significantly reduced; Worried about running out of traffic.
How did "Big Brother" come about?
MSU Video Codecs Comparison 2022
So, what are the key technical points of this intelligent code ultra-clear technology that is closely related to people's daily life? How does NetEase Yunxin do it?
To sum up, the key technical points of Smart Code ultra-clear transcoding are divided into three parts: video pre-processing, video encoding, and video post-processing .
The first is video pre-processing, including content analysis and image quality improvement. The first step of content analysis is to distinguish different scenes through scene recognition, such as: games, action movies, animation, video conferencing, etc., and select different pre-processing strategies and encoding tools according to the scene to achieve scene adaptation; Deep learning ROI area detection, the detected ROI area is passed to the pre-processing module and encoding module, and then the image quality of the ROI area is enhanced and repaired.
Image quality improvement starts from three aspects: video enhancement, color enhancement, and video noise reduction. The first two focus on improving the subjective image quality by enhancing the original video, while video noise reduction is to perform noise reduction processing after noise evaluation to improve subjective image quality. While improving the image quality, it can also help the compression rate of video encoding.
Next is video encoding. Intelligent coding includes perceptual coding, ROI coding, and precise frame-level and row-level code control; the coding core is completed through self-developed NE264, NE265, and NEVC that supports proprietary protocols.
In terms of video post-processing, the image quality is mainly improved through video super-resolution and video enhancement technologies. Video super-resolution refers to the conversion of low-resolution images into high-resolution images through the NetEase Yunxin AI super-resolution algorithm. NetEase Yunxin proposed a video super-resolution technology based on encoding loss recovery, using a strategy of data-driven and network design, simulating real distortion scenes through data processing, and optimizing layers from model design to engineering implementation. It has achieved good results in terms of super-resolution effects with real scenes.
media 1
Save bandwidth costs by more than 60%
MSU Video Codecs Comparison 2022
It is worth mentioning that this is not the first time that Netease Yunxin Audio and Video Lab has shined on the world stage.
In 2022, at the CVPR (CVPR NTIRE 2022 Efficient Super-Resolution Challenge), the highest-level conference in the field of computer vision, the AI super-resolution technology of Netease Yunxin Audio and Video Lab also won the world's first comprehensive performance track.
It is better to start and act than to sit and talk. Technology accumulation needs to be implemented in real scenarios to reflect its value. Within NetEase Group, NetEase Yunxin Zhicode ultra-clear technology has been fully launched in NetEase Media, NetEase Cloud Music, NetEase Yuanqi and other businesses.
Taking Netease Media as an example, according to last year's statistical results, the daily bandwidth of 400,000 minutes of video transcoding has dropped from 80G per day to less than 32G per day, saving more than 60% of bandwidth costs.
In the future, NetEase Yunxin will continue to explore cutting-edge technologies and provide the industry with lower-cost, lower-latency, and higher-quality video encoding and compression solutions.