Writing in the front

The 2018 World Cup in Russia is in full swing. With the entry of Internet video platforms and the popularity of new media live matches, this World Cup also has a new label: The Handheld World Cup. This means that users will be able to “lie down and watch the ball” more freely.

Half of the fans chose Youku and over 60% said it was hd and smooth

More than 50 percent of mobile phone users who watched the World Cup this year chose Youku, according to a survey by Streammedia. However, when new media platforms are added to live broadcast, most users are satisfied with the clarity and fluency of live broadcast.

More than 8 million users watched the World Cup on smart TVS during the World Cup, turning the living room into the home of the tournament, according to another figure. Only real hd picture quality, in order to withstand the double test of mobile phone ultra clear screen and TV ultra large screen.

When you watch a video, if the network is not good, there will be a lag or the system prompts to switch to standard definition or smooth quality, this time you will find that the video clarity is reduced. Hd means high resolution and requires more bandwidth to support. Hd and broadband are contradictory. This Session of Youku World Cup live received unanimous praise of the key, Ali Cloud “narrow band HD 2.0” played a significant role.

So what exactly is narrowband HD?

Bandwidth cost is a very important infrastructure cost in video service. How to reduce the cost under the premise of ensuring video quality is a crucial link in the whole link. Therefore, in video service, video encoding and decoding are very important technologies.

The transcoding technology in the industry has changed from MPEG2, to H.264, to H.265, which is roughly shown in the technical development curve below. Every ten years, the video compression rate will be doubled. On average, the industry video compression rate will be increased by less than 7% per year. Under this objective law of development, it has become very difficult for practitioners in the video industry to create overwhelming competitive advantages for their competitors.

Traditional video coding aims at minimizing signal loss, and strives to minimize the difference between output video and input video. However, in fact, the same coding loss may vary greatly in the definition of human eyes, because human eyes are sensitive to different picture positions. Narrowband HD is a set of video coding technology based on the optimal subjective perception of human eyes. Through intelligent analysis of live broadcast content, dynamic processing and coding are carried out according to the sensitivity of human eyes to achieve the best subjective sharpness effect.

There are three sets of visual models behind Narrowband HD 2.0

A model of the relationship between fidelity and subjective perception

As the fidelity of the video gets higher and higher, the human eye gradually loses the sense of it, so coding is stuck at the critical point where the distortion does not change much, and you can save bandwidth appropriately. The graph below shows the relationship between fidelity and subjective feeling model.

A relational model of resolution and bit rate

In order to guarantee smooth under the condition of a network to broadcast, video rate needs to have a limit, within the limit condition, if the resolution selection is too high, can lead to the assigned to each pixel is too little, video fuzzy, if choose low resolution, although can describe clearly each pixel, but need to rely on players do interpolation amplification, video will be blurred. Narrowband HD combines the analysis of a large number of video samples. Under a certain bit rate setting, it intelligently selects the most appropriate resolution according to the content of the video to achieve the best subjective visual effect.

Visual sensitivity model

Narrowband HD divides video content into parts that the human eye tends to pay attention to and parts that the human eye tends to ignore, and treats those parts differently. For the part that human eyes are easy to pay attention to, apart from the focus area, human eyes also pay attention to the texture of rules, such as the line of the penalty area in the court and the outline of the players. These are the areas that we must protect. We will make some adjustments and optimizations in these areas to make them more prominent and make the picture more tension. For the parts that are easily overlooked by the human eye, such as the defocus area, we can save the processing of this area. At the same time, we can also save the bandwidth by removing some small details that have no aggregation effect. For example, the background of the spectators’ seats in the stadium or the billboard is an area where the eyes are out of focus or a small detail with no aggregation effect. In addition, the human eye will also be very disgusted with burrs, mosaics, continuous flashing, we will deal with these details more gentle, clear, can improve the overall picture.

Below for narrow-band hd for the increase of the human eye focus area for messi hair set pieces of the scene in the figure, image focused on Lionel messi’s face, in without narrowband Gao Qingshi (left), will find macy’s face many details are missing, see picture is more smooth, played a bit like a layer of powder, and through narrowband hd encoding (right), The details on the face are completely retained and appropriately strengthened, so that the picture has a strong three-dimensional sense.

Narrowband HD World Cup upgrade

After the introduction of the above three models, in the live broadcast of this World Cup, special coding strategies were optimized based on machine learning for football live scenes. For example, football, grass and players were optimized by special coding strategies respectively, which greatly improved the sense of layers and transparency of the match pictures.

In football matches, the pictures often switch between different scenes, so how to keep the clarity in the scene switch requires a sophisticated bit rate control strategy. On the one hand, the video bit rate cannot be constant. If it is constant, the sudden scene switch will cause the bit rate allocated to the moving scene to be too small, resulting in blurred picture. On the other hand, the video bit rate should not fluctuate greatly. If it does, it will easily lead to user shortage and CDN bandwidth drastic fluctuation. Such drastic fluctuation will be a great challenge to CDN scheduling. Analyzing the motion intensity of narrowband HD video, combining the statistical learning of long-term bit rate allocation of video and the complexity of real-time video motion, bit rate shifting within the limited buffer range of player can improve subjective clarity as much as possible while ensuring smooth playback under the specified bandwidth.

Below is another World Cup broadcast (left) and Youku (right) at 1080P and 480P. When switching to this rich and dynamic corner fight, the difference between narrowband HD and non-narrowband HD is huge. When narrowband HD is not used, the bit rate allocation for this scenario is severely underrated, resulting in a very poor user experience.

The figure below shows the evaluation comparison between narrowband HD and X264 and X265 at different bit rates. It can be seen that there is a gap between narrowband HD and ordinary encoder in both subjective evaluation and VMAF score, which can save 30% bandwidth under the same definition.

Write in the last

As a strong technical support, Ali Cloud and Youku have solved the pain points of users with slow and low definition of sports events, and captured a large number of fans with the ultimate viewing experience of high-definition, smooth and zero delay. Of course, in addition to narrowband HD technology, Ali Video Cloud also has many industry-leading technologies, and has become the largest cloud computing company in terms of video service volume in China. Since the birth of Ali Cloud video cloud, has been committed to using their own technology, to create something unique in the industry. Ali Cloud will build a different video cloud service through ali Group’s years of technology precipitation, so that customers become different.

The original link