A key part of ultra-high-definition video.
01 Ranked first in many major indicators, saving 48% of bandwidth
Recently, the full HD 10bit track results of the 17th MSU World Encoder Competition were announced. Ali’s self-developed H.266/VVC encoder Ali266 won the most efficient 1fps grade on the track Two champions, compared with the competition benchmark encoder x265, it can save 48% of bandwidth, Effectively lower the threshold for ultra-high-definition video and promote its popularity.
The MSU World Coder Competition refers to a coding competition hosted by Lomonosov Moscow State University (MSU) for major companies, academic institutions, open source communities and individuals around the world. It has been held annually since 2005 and has been held so far. 17th session.
In the past 17 competitions, the overall number of participating encoders has continued to rise, making the MSU World Encoder Competition the most influential and top-level authoritative event in the field of video encoding and decoding, attracting participation from many well-known technology companies such as Google, Intel, and Netflix. The participating encoders all have broad practicality and represent the vane of industry development.
A total of 21 encoders participated in this MSU competition, and a 10-bit track was specially set up for the first time. After winning 8 championships in the 16th full HD track and the subjective track championship, Ali266 participated in the 1fps speed gear with the highest compression rate of the 10bit track in this competition, and achieved both SSIM and PSNR. No. 1 in this indicator.
In order to comprehensively evaluate the compression performance of participating encoders under multiple quality evaluation indicators, the MSU competition usedPSNR, SSIM, and VMAF and other objective quality evaluation indicators.
Among them, the SSIM index estimates the visual quality of the distorted image from three aspects:brightness, contrast and structural information, aiming to compare the original video Similarity to the structure of distorted videos, studying the damage of perceptual structures to evaluate video quality can better reflect the subjective characteristics of the human eye, so it has always been MSU The main evaluation indicators specified by the organizer.
Specifically,ranked by YUV (6:1:1)-SSIM indicator:
First place: Alibaba Ali266, Tencent Tencent266 v0.2.1 and Tencent266 v0.2.2
In the figure above, the ordinate is the participating encoders, and the abscissa is the average size of the output files of each encoder relative to the reference encoder x265 under the same SSIM quality.
The shorter the histogram is, the smaller the file output by the encoder is, the higher the compression rate is, and the better the encoder performance is. The figure shows that under the same YUV (6:1:1)-SSIM image quality, Ali266 saves 48% of files compared to the reference encoder x265 size.
Ranked by YUV (6:1:1)-PSNR (avg. MSE) indicator:
First place: Tencent Tencent266 v0.2.1, Alibaba Ali266
The figure shows that under the same YUV (6:1:1)-PSNR (avg. MSE) image quality, Ali266 saves 43%than the reference encoder x265. The file size of /span>.
Comprehensive, fair, and unbiased reviews of the many coders around the world are no easy task. Taking this year's 10bit track evaluation as an example, it took about 16 months from the public collection of participating encoders on June 1, 2022, to the release of the evaluation results on September 25, 2023.
Behind the time-consuming and laborious process lies the fundamental and critical role of video coding technology in the transmission and processing of multimedia information.
02 4K, 60 frames, 10bit, completing the last link of the complete link
Digital video is essentially a continuous frame of images. Although the size of a frame of image is not large, there must generally be at least 24 frames of images per second, and they will occupy a very large space when accumulated.
At present, the trend of ultra-high definition video is unstoppable, and people are increasingly enjoying the ultimate shocking experience brought by high resolution, high frame rate, and high bit depth of video.
Taking 4K ultra-high definition video as an example, with a resolution of 3840×2160 pixels and a frame rate of 60 (that is, 60 images per second), the data volume of an uncompressed 1-second video exceeds 11.94 billion bits (3840×2160 pixels). /frame×24bits/pixel×60 frames/second).
The video encoder can remove redundant information from the original video to "slim down" the video. Take the summer hit "Fengshen Part 1: Chao Ge Fengyun" as an example. The film is 148 minutes long, totaling 8,900 seconds. If the highest high-definition image quality of 4K, 24 frames/second, and 10 bit depth is selected, the entire film's data volume will exceed 7,000GB. With such a huge amount of data, it is almost impossible to directly transmit and store it without compression.
Under the premise of ensuring image quality, the encoder can compress the data amount of the original video to a few hundredths or even a few thousandths.
Therefore, video encoding technology makes the storage and playback of videos possible.
According to estimates, the full 4K file size of "Feng Shen" encoded using the x265 veryslow gear of the previous generation standard open source encoder that is widely used is about 3GB. However, using the Ali266 slow gear can save 1.8GB of traffic compared to the original solution, and the code rate can be saved up to 64 %.
There is no doubt that 4K, 60 frames, and 10 bit have gradually become the industry-recognized ultra-high-definition video standards. This year's MSU World Encoder Competition has set up a 10-bit track for the first time. So, what exactly can 10bit bring to our lives?
If you describe it in the most intuitive way, 10bit can make colors show more delicate gradient changes, because the color level in each color channel is 256 levels from 8bit (most display devices currently have (using 8bit) is suddenly increased to 10bit level 1024, which means that it canshow extraordinary delicacy in terms of color gradients and changes.
In the picture of the setting sun shown below, the upper half of the picture is represented by 8-bit bit depth. We can see that there is an obvious "fault" phenomenon in the orange-yellow transition of the sun from the inside to the outside, while the lower half of the picture is represented by 10-bit, and the color transition is very nature.
This comparison picture comes from the Internet
As national policies continue to develop and market demand continues to grow, the video ultra-high definition industry is booming. The high resolution, high frame rate, wide color gamut, wide dynamic range and other characteristics of ultra-high-definition video must be matched with high bit depth to fully display color fineness and contrast, and bring consumers a true live video experience. Therefore,high bit depth is one of the indispensable and important features and trends of ultra-high definition video.
To truly enjoy 10-bit color, it is not enough to have a screen that supports 10-bit display. Instead, the entire link of video collection, processing, encoding, storage or transmission, decoding, and display must be processed in 10-bit.
It can be seen that 10bit encoding and 10bit decoding are two essential links in the entire processing link. In the previous generations of H.266/VVC standards, the entry-level level only supported 8-bit bit depth, and 10-bit bit depth needed to be supported at higher expansion levels, so most codec devices did not 10bit bit depth video is not supported.
When H.266/VVC issued a technical solicitation in October 2017, it regarded wide color gamut and wide dynamic range video (i.e. HDR/WCG) as its main applications. Therefore, its entry level (Main10 Profile) supports The 10-bit bit-depth encoding standard greatly improves the friendliness of 10-bit bit-depth videos and aligns the video industry with supporting high bit-depth levels.
This time the Ali266 encoder won the award in the 10-bit track, which also proves that Ali266 is fully capable of 10-bit encoding and completes a key link of 10-bit in the full link. It is in line with the development trend of video ultra-high-definition technology and provides consumers with a true on-site video experience. New solutions are provided.
03 Continuous hard work, Ali266’s self-evolution
Ali266 is Alibaba Damo Academy’s codec implementation of the new generation international video standard H.266/VVC. It has high compression performance, high-definition real-time encoding speed, and complete real-time High-definition encoding and decoding capabilities and other features. The launch of the Ali266 codec has better opened up the end-to-end ecosystem of the H.266/VVC standard and provided the industry with a new generation of video coding and decoding solutions.
On theencoding end, Ali266 improves encoding quality and compression efficiency by implementing a variety of encoding algorithms, such as motion compensation time domain filtering, automatic Adapt to GOP size decision-making, scene switching detection, screen content detection, code rate control technology, etc.
On the other hand, Ali266 covers hundreds of fast encoding algorithms, and cooperates with engineering optimization methods such as multi-threading technology, assembly instructions, and memory access efficiency to greatly increase encoding speed at a minimal cost of compression performance.
On thedecoding end, Ali266’s self-developed decoder architecture, data structure and memory reusable design are optimized through multi-core parallelism and assembly , memory usage and memory access efficiency optimization and a series of engineering and algorithm optimization methods to improve decoding speed, and can be perfectly compatible with Android, iOS, Linux, MacOS, Windows and other platforms. Especially for mid-to-low-end mobile phones that integrate Ali266 decoders, they can also watch the latest H.266/VVC videos or live broadcasts clearly and smoothly, which better meets the needs of users in the mobile Internet era.
In addition, Ali266 fully considers the needs ofcommercialization software. After large-scale testing of thousands of high-end, high-end and low-end devices on different platforms, The robustness, stability and commercial availability of the Ali266 decoder are verified. Winning consecutive awards at the MSU World Encoder Competition marks that Ali266 has industry-leading software encoding and decoding performance and demonstrates its application potential in the ultra-high-definition video industry.
In January 2022, Ali266 was officially launched on Youku, and Youku became the industry’s first practical H.266/VVC implementation project at that time. According to estimates, since Youku has stably launched the use of Ali266, the bit rate has been saved by up to 40%40% compared to the original H.265/HEVC solution with the same picture clarity. ; In terms of experience, the lag rate is reduced by 50%, and the stability exceeds 99.95%< /span>.
In order to fully unleash the technical dividends brought by the upgrade of video codec standards and provide the industry with lower-cost, higher-quality video solutions, Alibaba Cloud and DAMO Academy have implemented Ali266's full support for video on demand services. This move will further help customers significantly save bandwidth costs, improve playback experience, and resolve the conflict between video viewing experience and bandwidth traffic.
Alibaba Cloud video on demand supports H.266/VVC, which mainly includes two aspects. On the one hand, Alibaba Cloud Video Cloud supports transcoding videos into H.266/VVC video streams, and supports mainstream containers such as mp4, ts, and hls to facilitate the storage, transmission, and distribution of H.266/VVC video streams. On the other hand, Alibaba Cloud Player provides a playback solution that is perfectly compatible with H.266/VVC encoding protocol video streams, allowing customers to enjoy a smooth and clear playback experience.
For more coding effect display, please click:https://retina.aliyun.com/#/Ali266
In the future, Ali266 will closely follow the latest technological development trends such as 10bit HDR, ultra-high definition 4K~8K, high frame rate 60fps~120fps, and free viewing angles, closely integrate with the audio and video industry, and continue to explore new businesses in on-demand, live broadcast, RTC and other scenarios. Possible applications, and deep integration with 5G, artificial intelligence, virtual reality and other technologies, spawning a large number of new scenarios, new applications, and new models, bringing audiences a more extreme audio-visual experience and more innovative interactive gameplay.